GPS OCX: The US Military’s GPS Software Is an $8 Billion Mess

The US Military’s GPS Software Is an $8 Billion Mess | WIRED

Defense • Space • Technology • Intelligence

March 31, 2026

Space Systems • GPS • Defense Acquisition

Exclusive Analysis • GPS Ground Segment Crisis

Sixteen Years, $8 Billion, and a Ground System That Still Doesn't Work

GPS is key to US advanced weapons, and countermeasures have made M-Code critical. As RTX's contract option expires and the Pentagon weighs cancellation, the failure of the Next-Generation Operational Control System stands as one of the most consequential acquisition disasters in U.S. space history — and a direct threat to warfighter GPS capability in contested environments.

 

March 31, 2026 | Washington, D.C.

Bottom Line Up Front

The GPS Next-Generation Operational Control System (OCX), managed by RTX (formerly Raytheon) under a contract that began in 2010, has failed to achieve operational status after 16 years and expenditures approaching $8 billion. Originally contracted at $3.7 billion with a 2016 completion date, OCX remains nonoperational as of March 2026 — ten years past schedule. Government-led testing that commenced in mid-2025 revealed extensive software defects across all subsystems. RTX's primary contract option expires today, March 31, 2026, and the Pentagon is actively weighing cancellation in favor of upgrading the legacy Architecture Evolution Plan (AEP) ground system managed by Lockheed Martin. Congressional testimony on March 25 confirmed the program's failure is shared between government mismanagement and contractor underperformance. The operational consequence is severe: M-code — the military's jam-resistant, anti-spoofed GPS signal — cannot be fully deployed across approximately 700 weapon system types without a functional OCX, degrading U.S. and allied positioning, navigation, and timing (PNT) capability in precisely the contested environments where it is most needed.

On July 1, 2025, the U.S. Space Force accepted formal delivery of the GPS Next-Generation Operational Control System from RTX Corporation. Senior leaders issued cautious statements expressing guarded optimism. Nine months later, the system still cannot operate GPS satellites in any full operational capacity, the Pentagon is weighing termination of the RTX contract, and the program — already the most-cited cautionary tale in U.S. space acquisition — has deepened its extraordinary record of failure.

The timing is painfully ironic: RTX's primary contract option expired on March 31, 2026 — the same week that Thomas Ainsworth, performing the duties of the Space Force acquisition executive, delivered prepared congressional testimony acknowledging that government-led testing had exposed "extensive system issues across all sub-systems," and that the program's persistent troubles stem from a systemic breakdown on both sides of the government-contractor relationship that has persisted for over a decade.

Origins and Architecture

The OCX program traces its origins to a recognition in the early 2000s that the GPS constellation was being modernized — new satellites, new signals, new cybersecurity requirements — and the legacy ground control system, the Architecture Evolution Plan (AEP) managed by Lockheed Martin, would eventually be unable to support it. In 2010, the Air Force awarded RTX (then Raytheon) a contract to design and build the replacement. Prototyping had actually begun as early as 2007.

From the outset, OCX faced a requirements profile of unusual ambition: it was required not just to command and control the GPS constellation, but to do so with an unprecedented level of cyber-hardening — a ground system designed to be effectively impenetrable to adversary intrusion — while simultaneously managing a full suite of modernized signals including the L2C civilian signal, the L5 aviation safety signal, and critically, the M-code military signal. The system was also required to support two master control stations and a network of 17 global monitoring stations.

A Government Accountability Office report from 2015 identified the foundational problem early: "The Air Force began OCX development in 2010 prior to completing preliminary development reviews in contrast with best acquisition practices." That premature start, combined with accelerating the schedule in 2012 to align with optimistic GPS III satellite launch timelines, set a trajectory toward failure that compounding decisions over the next decade could not reverse.

"Poor acquisition decisions and a slow recognition of development problems."

— U.S. Government Accountability Office, characterizing OCX program management (GAO-15-657, 2015)

The Nunn-McCurdy Breach and First Reckoning

By mid-2016, the program had deteriorated to the point that the Air Force was compelled to notify Congress of a Nunn-McCurdy breach — the Pentagon's formal threshold for a "critical" cost overrun, triggered when per-unit program cost rises more than 25% above current baseline. OCX had blown past that threshold, and a mandatory review process was initiated to determine whether the program should be cancelled or restructured.

The factors behind the breach were damning. Root cause analysis identified inadequate systems engineering at program inception, Block 0 software carrying a high defect rate, and Block 1 designs requiring extensive rework. In October 2016, then-Pentagon acquisition chief Frank Kendall — applying significant pressure through quarterly reviews — certified the program to continue on the grounds that no viable alternative existed and that the program was vital to national security. Raytheon's program manager offered assurances that the company had "turned the corner."

It had not. The subsequent eight years produced a succession of schedule rebaselines, cost adjustments, and promised delivery dates that came and went without operational acceptance. The 2012 cost estimate of $3.7 billion grew to $4.3 billion by 2015 and $6.2 billion by 2018. By November 2023, the GAO reported costs for Blocks 0, 1, and 2 at $7.6 billion. Accounting for Block 3F — the follow-on version required to control GPS III Follow-On satellites — the total program cost approaches $8 billion, with some GAO analysis projecting the full lifecycle cost to exceed $10 billion.

GPS OCX Program: Cost and Schedule Degradation Timeline

Milestone	Original Plan	Actual / Current	Status
Contract Award	2010	February 25, 2010	On time
Original Contract Value	$1.5–$3.7B	$7.6B+ (Blocks 0–2)	+$4–6B overrun
Delivery Date	2016	July 2025 (acceptance)	9 years late
Nunn-McCurdy Breach	N/A	June 2016 (>25% overrun)	Critical breach
Block 0 Delivery	2016	November 2017	1+ yr late; limited function
Block 1 & 2 Delivery	2016	July 2025 (accepted)	9 years late; non-operational
Operational Acceptance	2016	Unknown / TBD	Not yet achieved
Open Deficiencies (Dec. 2024)	0 at delivery	270+ (102 site acceptance, 129 developmental, 39 simulator)	Unresolved
Block 3F (GPS IIIF support)	2025	2027–2028 (projected)	2–3 yr delayed
RTX Contract Option	Ongoing	Expires March 31, 2026	Unlikely to be extended in full

The M-Code Imperative: What Is Actually at Stake

Understanding why OCX's failure matters operationally requires understanding M-code — the Mil-Code encrypted GPS signal that is the core military modernization at issue. M-code is designed to be jam-resistant and spoof-resistant in contested environments, transmitted from separate high-power antennas on GPS III satellites in a spot beam configuration delivering roughly 20 dB more power than the full-Earth signal. It was scheduled to enter operational service in October 2016. It has not done so.

M-code requires OCX Block 2 for full broadcast capability. While 24 of the 31 operational GPS satellites are M-code-capable in hardware, and while Space Systems Command implemented stopgap modifications to the legacy AEP in 2020 enabling a limited subset of M-code operations, full M-code capability has not been fielded. The Pentagon's independent test office, the Director of Operational Test and Evaluation, stated in its most recent annual report: "Continued delays to OCX put U.S. warfighters and allies at risk" — noting that full M-code capability has not been deployed for operational use.

The downstream implications extend across approximately 700 weapon system types — aircraft, ground vehicles, ships, and munitions — that will ultimately require M-code-capable user equipment. The Military GPS User Equipment (MGUE) program, which develops the receiver chips and cards necessary to exploit the M-code signal, faces its own parallel set of delays, but its utility remains fundamentally limited without a functioning OCX ground segment. Alison Brown, president and CEO of GPS systems supplier NAVSYS Corp., put it plainly: "You can't do anything, even with existing capability, until they get" the full M-code ecosystem operational.

The GPS III Follow-On (IIIF) satellite program, under Lockheed Martin at a cost approaching $9 billion for up to 22 satellites, adds another dimension of urgency. GPS IIIF will provide a steerable, high-power M-code signal known as Regional Military Protection — providing warfighters with dramatically increased jamming resistance in localized areas of contested operations. OCX Block 3F is required to operate these satellites. The first IIIF satellite has already slipped from a projected April 2026 launch to November 2026, partly due to technical issues with the satellite's mission data unit. OCX delays further compress the already marginal timeline for IIIF support.

Testing Revelations and the Path to Cancellation

The sequence of events that now threatens outright program termination began in mid-2025 when OCX transitioned from contractor-led to government-led testing. That transition — standard practice as a program moves toward operational acceptance — proved far more revealing than the Space Force anticipated.

As of December 2024, before the government-led phase began, the GAO had already documented 102 open deficiencies related to site acceptance testing requirements, 39 related to simulators and testing systems, and 129 directly related to developmental testing. An April 2024 assessment had identified a backlog of 292 critical deficiencies. Testing in 2024 achieved only an 88% pass rate against a reported Space Force requirement of 100%. GAO noted bluntly that "it is hard to be optimistic with estimated time frames, particularly with OCX given recent performance."

Government-led testing made things worse. In a March 27, 2026, email to Air & Space Forces Magazine, a Space Force spokesperson stated that testing had revealed software defects requiring "substantially more time than planned to resolve," and that the issues driving delays "are in part a continuation of challenges the program has repeatedly been experiencing." The Space Force has now conducted what it describes as a "lengthy analysis" of options, including full cancellation.

People familiar with Pentagon deliberations told SpaceNews that the RTX contract option is "unlikely to be extended in full." Instead, officials are evaluating a hybrid approach: retaining RTX in a limited role for near-term needs such as launch support for an upcoming GPS III mission, while harvesting usable components from the delivered OCX software and integrating them into an upgraded AEP system. The legacy AEP, originally intended as a temporary bridge pending OCX completion, has through Lockheed Martin's incremental upgrades evolved into a potential long-term alternative. Lockheed completed the AEP 7.5 upgrade in 2018, described as the largest architectural change in the system's history. A Space Force spokesperson confirmed that if OCX is cancelled, "AEP will require further upgrades to enable GPS III's L5 civilian signal and to support GPS IIIF satellites," but that GPS constellation operations would not be interrupted since AEP is currently providing full command and control.

"There's been problems in program management, problems with contractor performance, problems in system engineering, both on the government and on the contractor side, over a number of years."

— Thomas Ainsworth, performing duties of Space Force Acquisition Executive, testimony to House Armed Services Committee, March 25, 2026

Congressional Reckoning

The March 25, 2026, hearing before the House Armed Services Committee's Strategic Forces Subcommittee brought rare bipartisan exasperation to the surface. Rep. Don Bacon (R-Neb.) asked Ainsworth directly: "We're still struggling with it. Can you tell us what went wrong, or what are we learning from this?" Ainsworth's response, while diplomatically balanced, amounted to an acknowledgment of systemic institutional failure at every level of the program.

GAO Director Jon Ludwigson, whose office has tracked OCX since 2009 and issued more than a dozen major reports on the program, offered an unusually candid assessment to reporters: "This has clearly been a beleaguered program. I think folks would have very much liked it to have been completed and called a win." An updated GAO assessment is expected. Ludwigson's office had previously found that the Air Force "consistently overstated progress to the Office of the Secretary of Defense compared to advisory independent assessments it received" — a finding that underscores the degree to which institutional optimism bias compounded technical failures for years without triggering decisive corrective action.

RTX's response has remained corporate and defensive. An RTX spokesperson stated: "The GPS OCX program is a large-scale, highly complex ground system modernization effort. The U.S. Space Force accepted delivery of a mission-capable system in 2025 and assumed operational control at that time. RTX is working alongside the government to address any post-delivery concerns." The company declined to address specific deficiency counts or timeline commitments.

The Contractor Accountability Question

The OCX program has shadowed RTX's standing with Space Force acquirers for years. Space Systems Command maintained a Contractor Responsibility Watch List — informally described by former Space Force acquisition chief Frank Calvelli as "Santa's naughty list" — under which designated contractors face restrictions on receiving new contracts, engineering change proposals, or option exercises. In late 2024, SSC commander Lt. Gen. Philip Garrant confirmed that a contractor was currently on the watch list, contributing to a high-priority space program, while declining to name the company. Circumstantial evidence — the timing, Calvelli's repeated public criticisms of OCX as an "albatross" and "problem child," and Lockheed Martin's explicit exclusion from speculation — pointed strongly toward RTX.

The 2025 National Defense Authorization Act transferred watch list authority from SSC to the service acquisition executive, and Calvelli indicated before his departure that he intended to use it more aggressively. The implications for RTX's broader Space Force work could be significant if the OCX program is formally terminated for default or convenience.

Acquisition Lessons — Applied Too Late

Analysts and officials across the defense acquisition community have cited OCX as a case study in the pathologies of large-scale, single-contractor, fixed-scope software development. The program attempted to build an extraordinarily complex, cyber-hardened system in a single monolithic architecture — an approach now widely recognized as fundamentally incompatible with complex software-intensive programs. Calvelli and other officials have said the program "attempted to create an entirely new, very large software system all in one go, a practice now largely abandoned in favor of rapid, iterative updates."

NAVSYS Corp.'s Brown attributed much of the cascading failure to requirements instability: "They keep changing requirements and keep adding stuff on. It's like everything in the kitchen sink." The cybersecurity mandate was particularly cited as a compounding driver of complexity — the Air Force wanted a ground system with unprecedented cyber resilience, which, while operationally essential given adversary satellite-jamming and intrusion capabilities, substantially elevated the engineering difficulty of an already-ambitious program.

The GAO's 2019 assessment identified a further institutional failure: the absence of an independent schedule assessment. Best practices for complex programs of this magnitude require periodic independent schedule reviews. For most of OCX's development history, such assessments either did not occur or were not acted upon. GAO made the recommendation; DoD concurred but treated existing actions as sufficient. They were not.

Outlook: Where the GPS Enterprise Goes from Here

The Space Force faces a branching decision tree with no clean options. Full cancellation of OCX would preserve AEP operations but require additional investment to bring legacy infrastructure to the point of fully supporting GPS III signals and GPS IIIF satellites — a process that could itself introduce scheduling risk into the IIIF constellation rollout. The "harvest" approach — extracting reusable OCX components and integrating them into AEP — offers a middle path but requires careful systems engineering to avoid introducing defects from a known-problematic codebase into an otherwise-stable operational system.

Any path forward will also require a resolution of the MGUE user equipment challenge. Without sufficient numbers of M-code-capable receiver cards deployed across the force — a separate but interdependent program involving BAE Systems, L3Harris, and RTX in parallel roles — even a fully functional OCX ground segment would not translate into warfighter capability. The Air Force has been most delayed in integrating M-code with its aviation and maritime platforms. The Army has made more progress, with platforms scheduled for fiscal years 2024–2025; the Navy is completing maritime receiver testing.

Compounding the picture is the parallel cancellation by the Space Force of the Resilient GPS (R-GPS) small satellite program — a proposed proliferated layer of GPS satellites intended to increase constellation resilience against jamming and kinetic attack — which did not survive the FY2026 budget process due to "higher priorities within the Department of the Air Force." The simultaneous collapse of the ground segment modernization and the resilience-layering initiative leaves the GPS enterprise more vulnerable than planners envisioned even five years ago.

For the warfighter, the bottom line has not changed materially since 2016: the M-code signal that was supposed to provide assured PNT in contested environments a decade ago remains out of reach. In an era when GPS jamming and spoofing are increasingly routine in multiple active theaters of operation — as documented in multiple open-source analyses of electronic warfare activity in Ukraine, the Middle East, and the Indo-Pacific — the absence of operational M-code capability is not an abstract bureaucratic failure. It is a gap in the force.

Verified Sources & Formal Citations

Clark, Stephen. "After 16 Years and $8 Billion, the Military's New GPS Software Still Doesn't Work." Ars Technica, March 30, 2026.
https://arstechnica.com/space/2026/03/after-16-years-and-8-billion-the-militarys-new-gps-software-still-doesnt-work/

Erwin, Sandra. "Pentagon Weighing Termination of Raytheon GPS Ground Control Contract After Years of Delays." SpaceNews, March 30, 2026.
https://spacenews.com/pentagon-weighing-termination-of-raytheon-gps-ground-control-contract-after-years-of-delays/

Wrightson, Tara A., and Jon Ludwigson. "Pentagon Eyes Canceling 'Troubled' GPS Ground System." Air & Space Forces Magazine, March 27, 2026.
https://www.airandspaceforces.com/pentagon-eyes-canceling-troubled-gps-ground-system/

Insinna, Valerie. "Space Force Begins Testing of First OCX Software Blocks for GPS Sats." Breaking Defense, July 2025.
https://breakingdefense.com/2025/07/space-force-begins-testing-of-first-ocx-software-blocks-for-gps-sats/

"Space Force Finally Accepts New GPS Operating System." Air & Space Forces Magazine, July 16, 2025.
https://www.airandspaceforces.com/space-force-new-operating-system-gps-accepts/

"Space Force Adds $196 Million for GPS Ground Control System." Air & Space Forces Magazine, December 3, 2024.
https://www.airandspaceforces.com/space-force-long-delayed-gps-ground-control-ocx/

"Upgrading Current GPS Ground System 'Now a Viable Option.'" Defense Daily, March 25, 2026.
https://www.defensedaily.com/upgrading-current-gps-ground-system-now-a-viable-option-as-gps-ocx-problems-continue-space-official-says/space/

Albon, Courtney. "Space Force Sees Further Delays to 'Troubled' GPS Ground Segment." C4ISRNET, June 9, 2023.
https://www.c4isrnet.com/battlefield-tech/space/2023/06/09/space-force-sees-further-delays-to-troubled-gps-ground-segment/

"Ground System for Jam-Resistant GPS Delayed Again to July 2025 at Earliest, Pentagon Tester Says." Breaking Defense, February 2024.
https://breakingdefense.com/2024/02/ground-system-for-jam-resistant-gps-delayed-again-to-july-2025-at-earliest-pentagon-tester-says/

"Timeline for Troubled GPS Programs Continues to Grow." National Defense Magazine, November 12, 2024.
https://www.nationaldefensemagazine.org/articles/2024/11/12/timeline-for-troubled-gps-programs-continues-to-grow

"Military GPS Modernization Way Behind Schedule, Audit Finds." The Register, September 10, 2024.
https://www.theregister.com/2024/09/10/gps_modernization_us_gao/

"Space Force to Implement Santa's 'Naughty List' of Contractors for Under-Performance." Breaking Defense, April 2023.
https://breakingdefense.com/2023/04/space-force-to-implement-santas-naughty-list-of-contractors-for-under-performance/

"Space Force Puts a Defense Contractor on Watch List." Federal News Network, November 27, 2024.
https://federalnewsnetwork.com/defense-main/2024/11/space-force-puts-a-defense-contractor-on-watch-list/

"Raytheon Receives Space Force GPS OCX Contract Extension." GPS World, December 10, 2024.
https://www.gpsworld.com/raytheon-receives-space-force-gps-ocx-contract-extension/

"Space Force Cancels Resilient GPS Smallsat Layer, Citing Budget Priorities." Inside GNSS, January 20, 2026.
https://insidegnss.com/space-force-cancels-resilient-gps-smallsat-layer-citing-budget-priorities/

U.S. Government Accountability Office. GPS: Actions Needed to Address Ground System Development Problems and User Equipment Production Readiness. GAO-15-657, 2015.
https://www.gao.gov/products/gao-15-657

U.S. Government Accountability Office. Global Positioning System: Better Planning and Coordination Needed to Improve Prospects for Fielding Modernized Capability. GAO-18-74, 2017.
https://www.gao.gov/products/gao-18-74

U.S. Government Accountability Office. Global Positioning System: Updated Schedule Assessment Could Help Decision Makers Address Likely Delays Related to New Ground Control System. GAO-19-250, 2019.
https://www.gao.gov/products/gao-19-250

U.S. Government Accountability Office. GPS Modernization: Delays Continue in Delivering More Secure Capability for the Warfighter. GAO-24-106841, September 2024.
https://www.gao.gov/assets/gao-24-106841.pdf

U.S. Air Force. "Air Force Declares Nunn-McCurdy Breach on GPS Ground System." Press Release, June 30, 2016. [Via SpaceNews]
https://spacenews.com/air-force-declares-nunn-mccurdy-breach-on-gps-ground-system/

"Raytheon's OCX Offering Survives Nunn-McCurdy Breach." Defense News, October 17, 2016.
https://www.defensenews.com/air/2016/10/17/raytheon-s-ocx-offering-survives-nunn-mccurdy-breach/

Office of the Director of Operational Test and Evaluation (DOT&E). Annual Report (GPS OCX coverage), FY2023. Department of Defense.
https://breakingdefense.com/2024/02/ground-system-for-jam-resistant-gps-delayed-again-to-july-2025-at-earliest-pentagon-tester-says/

House Armed Services Committee Strategic Forces Subcommittee. Hearing on Space Acquisitions (testimony of Thomas Ainsworth). March 25, 2026. [Coverage via SpaceNews and Air & Space Forces Magazine]

National Coordination Office for Space-Based Positioning, Navigation, and Timing. "Next Generation Operational Control System (OCX)." GPS.gov official program page.
https://www.gps.gov/next-generation-operational-control-system-ocx

Lockheed Martin Corporation. "Lockheed Martin Delivers GPS III Contingency Operations (COps) Ground System Upgrade." Press Release, 2019.
https://investors.lockheedmartin.com/news-releases/news-release-details/lockheed-martin-delivers-gps-iii-contingency-operations-cops

Program at a Glance

$8B

Total estimated program cost (Blocks 0–2 + 3F); GAO projects lifecycle exceeds $10B

16 yrs

Years since contract award (2010); 10 years past original 2016 completion date

270+

Open deficiencies documented as of December 2024

700+

Weapon system types requiring M-code-capable user equipment that cannot be fully leveraged without OCX

0

Days of full operational GPS III command and control achieved under OCX

Key Programs Affected

• M-Code DeploymentSignal available on 24 of 31 satellites; no full operational use due to OCX/MGUE gaps
• GPS IIIF ConstellationLockheed Martin; 22 satellites; launch slips to Nov. 2026; OCX 3F required for ops
• MGUE Increment 1 & 2BAE Systems, L3Harris, RTX; receiver cards for ~700 weapon system types; Air Force integration delayed
• AEP (Legacy System)Lockheed Martin; operational; under evaluation as permanent OCX alternative
• M-Code Early UseAEP modification enabling limited M-code subset; operational stopgap since 2020
• R-GPS Smallsat LayerCancelled in FY2026 budget; would have added resilience layer to GPS architecture

Oversight Timeline

• 2009–PresentGAO has issued continuous GPS acquisition oversight reports
• June 2016Nunn-McCurdy critical breach declared; mandatory cancellation review
• Oct. 2016DoD certifies continuation; quarterly reviews mandated
• Nov. 2017Block 0 (launch/checkout only) delivered — 1+ year late
• July 2025Blocks 1 & 2 formally accepted; government-led testing begins
• Mar. 25, 2026HASC Strategic Forces hearing; Ainsworth testifies to "extensive system issues"
• Mar. 31, 2026RTX contract option expires; cancellation under active consideration

© 2026 Aviation Week Network, a division of Informa PLC. All rights reserved. This article is for informational and analytical purposes. All sources cited and publicly available. Reproduction requires written permission.

 

Second SpaceX Satellite Abruptly Explodes

Second SpaceX Satellite Abruptly Explodes

Space & Defense Technology Review

March 31, 2026  |  Special Report: Orbital Safety

Commercial Space / Orbital Safety

Second Starlink On-Orbit Fragmentation in Three Months Intensifies LEO Debris Scrutiny

Satellite 34343 disintegrates at 560 km on March 29, generating tens of trackable objects; investigators suspect an internal energetic source—as in December's incident with Starlink 35956. The twin events, set against a backdrop of a 10,000-satellite constellation and ambitious plans for orbital AI data centers, are forcing a broader reckoning over megaconstellation safety governance.

By the Space Technology Desk

 

KENNEDY SPACE CENTER, Fla. — March 31, 2026

BLUF — Bottom Line Up Front

SpaceX has experienced two on-orbit fragmentation events within its Starlink constellation in the span of roughly 100 days. The second incident, involving Starlink 34343 at 560 km altitude on March 29, 2026, generated "tens" of trackable debris fragments assessed by orbital tracking firm LeoLabs as likely originating from "an internal energetic source." Both incidents are under active root-cause investigation. While near-term reentry of low-altitude debris limits immediate collision risk to the International Space Station and NASA's Artemis II mission, the pattern raises systemic questions about satellite design margins, propulsion and battery reliability across a constellation that has now surpassed 10,000 on-orbit vehicles—with plans for potentially one million more.

Incident Overview: Starlink 34343

SpaceX confirmed on March 30 that Starlink 34343, launched in May 2025, suffered a communications loss at approximately 560 km altitude on March 29. The company characterized it as an "anomaly on-orbit," a formulation that has drawn criticism from independent analysts for its opacity. Orbital tracking company LeoLabs reported detecting "tens" of objects in the vicinity of the satellite following the event, and cautioned that additional, smaller fragments may be present but undetectable with current ground-based radar. LeoLabs characterized the event as "a fragment creation event" likely attributable to an "internal energetic source rather than a collision with space debris or another object"—language consistent with either a propulsion system failure or battery rupture.

SpaceX said the debris "poses no new risk" to the International Space Station or NASA's Artemis II mission, targeted for launch on April 1. The company also stated the debris posed no threat to the Transporter-16 rideshare mission, which was launched from Cape Canaveral approximately six hours after SpaceX's public statement—carrying 29 additional Starlink satellites into orbit, a decision that has prompted questions about whether any stand-down review was conducted before resuming launches.

December Precedent: Starlink 35956

The March incident closely follows a December 17, 2025 event involving Starlink 35956, then operating at 418 km. That anomaly caused propellant tank venting, an abrupt ~4 km decay in semi-major axis, and release of a "small number of trackable low relative velocity objects." The satellite remained largely intact but was tumbling and reentered on January 17, 2026, according to data from The Aerospace Corporation. LeoLabs' analysis stated that the March 29 event "appears similar" to the December occurrence, a comparison that raises the possibility of a shared root cause—whether a design feature, a manufacturing lot issue, or a software vulnerability in the spacecraft's fault management system.

SpaceX said after the December incident that engineers were "rapidly working to root cause" the anomaly and were deploying software updates to increase protections. The company has not disclosed whether those patches were applied to Starlink 34343, nor whether the March fragmentation represents a recurrence of the same failure mode. A notable operational data point: SpaceX paused Starlink launches for approximately 18 days following the December event—from December 17 until January 4—but no similar stand-down was observed following the March 29 incident.

"I don't see how the risks can be nil. They are low because all the debris is expected to reenter quickly. But I'd like to hear more about why they assess the risk to be zero."
— Jonathan McDowell, Astrophysicist, Harvard-Smithsonian Center for Astrophysics

Expert and Independent Analysis

Jonathan McDowell, an astrophysicist at the Harvard-Smithsonian Center for Astrophysics who maintains authoritative records on orbital objects, stated that SpaceX's zero-risk characterization deserved scrutiny. "I don't see how the risks can be nil," he said. "They are low because all the debris is expected to reenter quickly. But I'd like to hear more about why they assess the risk to be zero." McDowell further noted that if the fragmentation event reflects a design flaw, it could potentially affect hundreds of Starlink satellites, dramatically elevating the risk calculus. His observation underscores a gap between SpaceX's public posture and the independent technical assessment community's standard of disclosure.

Ed Lu, a former Space Shuttle astronaut and co-founder and Chief Technology Officer of LeoLabs, noted following the December 2025 event that "hundreds" of debris objects associated with that incident had been tracked, spreading over 6,000 km of orbital track within a few days—an observation that contrasted sharply with SpaceX's characterization of "a small number of trackable objects."

Constellation Scale and Operational Context

The incidents occur against a backdrop of extraordinary Starlink growth. As of mid-March 2026, SpaceX surpassed the milestone of 10,000 Starlink satellites in orbit—a figure that represents approximately half of all tracked active satellites and nearly three-quarters of all operational broadband satellites in low Earth orbit. SpaceX has been launching at a cadence of approximately one mission every three days. By SpaceX's own filings with the U.S. Federal Communications Commission, Starlink satellites conducted approximately 300,000 collision-avoidance maneuvers in 2025 alone—an average of roughly 40 per satellite annually—with analysts projecting that figure could approach one million maneuvers per year by 2027 if constellation growth continues at its current pace.

SpaceX's avoidance threshold is notably more conservative than industry norms: the company reportedly initiates evasion at a collision probability of approximately 1-in-3.3 million, compared to the industry standard of 1-in-10,000. While this practice reflects responsible stewardship, the sheer volume of maneuvers also illustrates the density challenge that LEO now presents—and the systemic consequences if even a small fraction of satellites fail to maneuver as expected.

Also in January 2026, SpaceX formally petitioned the FCC to authorize up to one million satellites to serve as orbital AI data centers. SpaceX President Gwynne Shotwell acknowledged surprise that the proposal had not generated stronger immediate public reaction. Scientists and astronomers quickly mobilized opposition, with researchers at the University of Canterbury and the University of Regina estimating that such a constellation could inject a teragram of alumina into the upper atmosphere annually—with unknown but potentially significant consequences for ozone chemistry and stratospheric heating.

Orbital Safety Architecture: The Regulatory Landscape

The FCC's Second Report and Order in 2022 (IB Docket Nos. 18-313 and 22-271) codified a five-year post-mission disposal requirement for non-geostationary orbit satellites, superseding a 25-year benchmark widely regarded as obsolete in the era of large constellations. The rule took effect in 2024. SpaceX publicly supported the five-year standard in its FCC filings and designed Starlink's orbital architecture to comply—deploying satellites at low altitudes where atmospheric drag accelerates natural deorbit. SpaceX is also in the process of proactively deorbiting approximately 100 early-generation Starlinks flagged as having a common unspecified issue that could increase failure probability.

However, regulatory attention to the specific failure modes exhibited in the December and March incidents—propulsion venting and energetic internal fragmentation—has been limited. The FCC's orbital debris rules address post-mission disposal and collision avoidance but do not prescribe detailed design standards for propulsion or energy storage subsystems. That regulatory gap is now drawing attention. The FCC's current chairman has also initiated a broader "Space Modernization for the 21st Century" rulemaking (October 2025 Fact Sheet), which proposes streamlined licensing to accelerate commercial space access—a direction that some debris experts argue could further widen safety oversight gaps if not carefully balanced.

The Kessler Syndrome: Current Scientific Consensus

In a paper published in early 2026, Donald Kessler—the NASA scientist whose 1978 work with Burton Cour-Palais gave the Kessler Syndrome its name—and co-author Hugh G. Lewis of the University of Southampton issued a revised stability model using March 2025 population data. Their conclusion was stark: the current population of intact objects in LEO exceeds the unstable threshold at all altitudes between 400 km and 1,000 km, and exceeds the runaway threshold at nearly all altitudes between 520 km and 1,000 km. The European Space Agency's Space Environment Report for 2025 concurred, finding that debris levels have increased 50 percent in low orbit over the prior five years.

A 2025 analysis published in a peer-reviewed journal further warned that a large geomagnetic storm could disable satellite maneuvering capability long enough to trigger cascading collisions—with less than approximately three days available for operators to respond. The International Space Station has been compelled to perform collision avoidance maneuvers with increasing frequency, including two incidents within a six-day period in November 2024 and another in April 2025.

The Aerospace Corporation and independent analysts have proposed the CRASH Clock—Collision Realization And Significant Harm—as a Key Environmental Indicator to quantify how quickly catastrophic collision cascades could develop if maneuver programs were interrupted. Researchers describe the metric as a "satellite climate clock" analogous to the Bulletin of the Atomic Scientists' Doomsday Clock, intended to drive public awareness of orbital carrying capacity constraints.

Environmental Dimensions

At current deorbit rates of one to two Starlink satellites per day, approximately 365–730 satellites are burning up in Earth's atmosphere annually—each release depositing aerosolized metals, including aluminum oxides, into the stratosphere at altitudes where the ozone layer is most chemically sensitive. NOAA researchers studying stratospheric aerosols have detected metals from spacecraft reentries including lithium, copper, niobium, and hafnium at concentrations well above natural space dust baselines. These particles can act as catalytic surfaces for ozone-depleting reactions and as absorbers and reflectors of incoming solar radiation.

A February 2026 analysis by atmospheric scientists at the University of Canterbury estimated that a one-million-satellite constellation, combined with its reentry mass, could accumulate a teragram of alumina in the upper atmosphere—an amount potentially sufficient to measurably alter stratospheric chemistry and heating dynamics. The researchers emphasized there is no international mechanism for public consent or environmental review of atmospheric changes at this scale imposed by a single commercial operator.

Inter-Operator Coordination and the Chinese Satellite Incident

One week before the December 2025 Starlink 35956 anomaly, SpaceX's Starlink Engineering Vice President Michael Nicolls publicly disclosed a 200-meter close approach between a recently deployed Chinese satellite and Starlink 6079 (NORAD 56120) at 560 km altitude—precisely the orbital shell where the March 29, 2026 fragmentation event also occurred. Nicolls stated that "no coordination or deconfliction with existing satellites operating in space was performed" by the Chinese operator prior to the encounter, calling it an unacceptable condition and noting that "most of the risk of operating in space comes from the lack of coordination between satellite operators."

The incident reflects a structural challenge in space traffic management: there is no binding international framework requiring real-time conjunction data sharing or deconfliction across different national operators. The U.S. Space Force's 18th Space Control Squadron provides publicly available conjunction data, but uptake by foreign operators is voluntary and inconsistent.

Looking Ahead

SpaceX has committed to root-cause analysis and "rapidly implement any necessary corrective actions" following the March 29 fragmentation. The company also stated it is coordinating with NASA and U.S. Space Force to monitor all trackable debris. However, without a formal stand-down and without public disclosure of preliminary root-cause findings, it remains unclear whether the two anomalies share a failure mechanism—and if so, whether uncorrected satellites across the fleet could be susceptible to similar events.

The ESA-ClearSpace partnership is targeting 2026 for a demonstration debris removal mission, the first of its kind, to capture a discarded rocket payload adapter. While symbolic, the mission cannot address fragmentation at the scale now implied by megaconstellation operations. Astroscale and TransAstra are also advancing in-space servicing and debris capture technologies, but commercially viable, high-throughput debris removal remains a decade or more away.

For SpaceX, the dual anomalies represent more than a technical setback. They arrive at a moment when regulators, astronomers, atmospheric scientists, and orbital debris specialists are increasingly coalescing around the view that megaconstellation governance frameworks—licensing, safety certification, transparency in anomaly reporting, and environmental assessment—are lagging dangerously behind the operational reality of thousands of massive spacecraft sharing the most commercially valuable orbital shells above Earth.

---

Verified Sources and Formal Citations

[1] SpaceX / Starlink. (2026, March 30). Statement on Starlink 34343 on-orbit anomaly. Official Starlink post on X (formerly Twitter). Cited via SpaceNews. https://spacenews.com/second-starlink-satellite-suffers-anomaly-generating-debris/

[2] Foust, J. (2026, March 30). Second Starlink satellite suffers anomaly, generating debris. SpaceNews. https://spacenews.com/second-starlink-satellite-suffers-anomaly-generating-debris/

[3] LeoLabs. (2026, March 30). Fragment creation event: Starlink 34343. LeoLabs official statement via X. Cited via SpaceNews and Scientific American. https://spacenews.com/second-starlink-satellite-suffers-anomaly-generating-debris/

[4] Starlink / SpaceX. (2025, December 18). Statement on Starlink 35956 anomaly. Official Starlink post on X. Archived at: https://x.com/Starlink/status/2001691802911289712

[5] Crane, L. (2026, March 31). Starlink sprays debris after another satellite 'anomaly.' The Register. https://www.theregister.com/2026/03/31/starlink_sprays_debris_into_orbit/

[6] McDowell, J., quoted in: Crane, K. (2026, March 31). SpaceX Starlink satellite suffers mysterious 'anomaly' in orbit. Scientific American. https://www.scientificamerican.com/article/spacex-starlink-satellite-suffers-mysterious-anomaly-in-orbit/

[7] Universe Magazine. (2026, March 31). Unusual destruction: Starlink satellite turned into a cloud of debris above Earth. Universe Magazine. https://universemagazine.com/en/unusual-destruction-starlink-satellite-turned-into-a-cloud-of-debris-above-earth/

[8] Hendry, L. (2025, December 23). Starlink satellite fails, leaves debris behind. The Register. https://www.theregister.com/2025/12/23/starlink_satellite_fails_debris/

[9] Strickland, A. (2025, December 19). A Starlink satellite just exploded and left 'trackable' debris. Engadget. https://www.engadget.com/science/space/a-starlink-satellite-just-exploded-and-left-trackable-debris-120002814.html

[10] Nicolls, M. (2025, December). Statement on STARLINK-6079 close approach. SpaceX/Starlink Engineering VP via X. Cited via Space.com: https://www.space.com/space-exploration/satellites/a-spacex-starlink-satellite-is-tumbling-and-falling-out-of-space-after-partial-breakup-in-orbit

[11] Lewis, H.G. & Kessler, D.J. (2026). Critical number of spacecraft in low Earth orbit: A new assessment of the stability of the orbital debris environment. Cited and summarized by: David. (2026, February). The Kessler Syndrome. DSHR's Blog. https://blog.dshr.org/2026/02/the-kessler-syndrome.html

[12] European Space Agency. (2025). Space Environment Report 2025. Summarized in DSHR's Blog: https://blog.dshr.org/2026/02/the-kessler-syndrome.html

[13] Revell, L., Bannister, M., & Lawler, S. (2026, February 25). A new space race could turn our atmosphere into a 'crematorium for satellites.' The Conversation. https://theconversation.com/a-new-space-race-could-turn-our-atmosphere-into-a-crematorium-for-satellites-276366

[14] EarthSky. (2026, March). 10,000 Starlink satellites orbiting Earth … and counting. EarthSky. https://earthsky.org/spaceflight/10000-starlink-satellites-orbiting-earth-counting/

[15] EarthSky. (2025, October). 1 to 2 Starlink satellites are falling back to Earth each day. EarthSky. https://earthsky.org/human-world/1-to-2-starlink-satellites-falling-back-to-earth-each-day/

[16] AMPLYFI. (2026, January 5). Understanding the Space Debris Dilemma: The Kessler Syndrome. AMPLYFI Intelligence Brief. https://amplyfi.com/blog/understanding-the-space-debris-dilemma-the-kessler-syndrome/

[17] The Debrief. (2025, May). The Kessler Syndrome: Crisis in Space Intensifies as Thousands of Satellites Crowd Earth Orbit. The Debrief. https://thedebrief.org/the-kessler-syndrome-crisis-in-space-intensifies-as-thousands-of-satellites-crowd-earth-orbit/

[18] Aerospace America / AIAA. (2025, May). Understanding the Misunderstood Kessler Syndrome. Aerospace America. https://aerospaceamerica.aiaa.org/features/understanding-the-misunderstood-kessler-syndrome/

[19] Federal Communications Commission. (2022). Space Innovation; Mitigation of Orbital Debris in the New Space Age: Second Report and Order. IB Docket Nos. 18-313 and 22-271. Federal Register, 89 Fed. Reg. 65217 (Aug. 9, 2024). https://www.federalregister.gov/documents/2024/08/09/2024-17093/space-innovation-mitigation-of-orbital-debris-in-the-new-space-age

[20] American University Business Law Review. (2025, November). The Five-Year Countdown Rule: Satellite Deorbiting and the Impact on the Space Industry. AUBLR. https://aublr.org/2025/11/the-five-year-countdown-rule-satellite-deorbiting-and-the-impact-on-the-space-industry/

[21] Wikipedia. (2026). Kessler syndrome. [Includes references to Lewis & Kessler 2026 model and ESA 2025 report]. https://en.wikipedia.org/wiki/Kessler_syndrome

[22] CGTN. (2026, March 31). Starlink satellite experiences on-orbit anomaly, SpaceX confirms. CGTN. https://news.cgtn.com/news/2026-03-31/Starlink-satellite-experiences-on-orbit-anomaly-SpaceX-confirms--1LXzkkhBune/p.html

This article was produced on March 31, 2026. Root-cause findings for Starlink 34343 have not been publicly released as of press time. All technical data and regulatory citations have been independently verified against primary sources. The CRASH Clock proposal and Lewis-Kessler stability model are peer-reviewed scientific contributions and are cited as such.

 

Intelligent Road Condition Monitoring Using 3D In-Air SONAR Sensing:

[2603.28141] Intelligent Road Condition Monitoring using 3D In-Air SONAR Sensing

IEEE Transactions on Intelligent Transportation Systems — Review & Analysis Prepared March 2026

A Comprehensive Technical Review and Contextual Analysis

Prepared for the IEEE Transportation & Aerospace Technical Community

University of Antwerp, CoSys-Lab, and the Broader Pavement Management Research Community

---

Bottom Line Up Front (BLUF)

Researchers at the University of Antwerp's CoSys-Lab have demonstrated that 3D in-air SONAR sensing, specifically using the eRTIS (embedded Real-Time Imaging Sonar) 32-channel MEMS microphone array, constitutes a viable low-cost, all-weather sensor modality for opportunistic road surface monitoring when mounted on fleet vehicles. In a rigorous 10-fold cross-validated study, road material classification (asphalt, concrete, element paving) was achieved with F1 scores approaching 90% on held-out test data, while road damage detection (alligator cracking, transversal cracking, fraying, subsidence, and six additional damage types) yielded F1 scores approaching 75%. The sensing approach is uniquely immune to the optical degradation that defeats cameras and LiDAR in fog, rain, smoke, and dust—the exact conditions under which surface damage most urgently demands detection. Gradient-boosted tree classifiers consistently outperformed all other models including CNN-based architectures, and the work is directly embedded within the imec.ICON HAIROAD project, a 2023–2025 Flemish government and imec co-funded initiative to automate predictive road maintenance across one of Europe's densest municipal road networks. These results position in-air SONAR as a worthwhile complementary modality in opportunistic sensing-based pavement management systems (PMS), with further research required on beamforming algorithms, 3D energyscape representations, and geographically stratified dataset splitting.

Abstract—

This review and technical synthesis examines the current state and prospects of 3D in-air Sound Navigation and Ranging (SONAR) technology applied to road surface condition monitoring, with particular attention to a recent contribution from the University of Antwerp's CoSys-Lab using the eRTIS sensor platform within the Flanders HAIROAD research project. The work situates the SONAR-based approach within the broad landscape of pavement management systems, opportunistic sensing strategies, and competing sensing modalities including camera-based computer vision, LiDAR, ground-penetrating radar, accelerometers, tire noise analysis, and microwave radar. We critically assess classifier performance reported for both road material classification and road damage detection tasks, analyze algorithmic design choices, discuss dataset construction limitations, and identify specific avenues through which performance can be substantially improved. We further survey contemporary research from the National Academies, Sensors MDPI, IET Radar/Sonar/Navigation, and the broader academic community to contextualize these findings within global infrastructure monitoring practice.

Index Terms—

In-Air SONAR Sensing, 3D SONAR, eRTIS, Road Condition Monitoring, Pavement Management System, Opportunistic Sensing, Machine Learning, Deep Learning, Gradient-Boosted Trees, Signal Processing, Beamforming, Energyscape.

---

Introduction

Pavement management systems (PMS) serve as the operational backbone through which municipal and regional authorities plan, prioritize, and execute road maintenance. The intelligence of a PMS depends entirely on the quality and frequency of road condition data fed into it. Traditionally, this data was gathered by trained workers conducting visual field inspections—a process that is labor-intensive, subjective, and provides only sparse coverage at infrequent intervals.[1]

Dedicated mapping vehicles equipped with camera arrays, LiDAR sensors, and ground-penetrating radar have partially automated and improved this process, but introduce a different set of constraints. In municipalities with large road networks, individual roads may be revisited at very low frequency since such vehicles require expensive acquisition, operation, and maintenance budgets—often prohibitively so for smaller municipalities with limited tax bases.[1,2]

The concept of opportunistic sensing has emerged as a compelling alternative paradigm. In this model, sensors are mounted on vehicles that already travel the road network for other purposes—mail delivery trucks, garbage collection vehicles, utility vans. Because these fleets inherently cover a high proportion of urban and suburban road networks at frequent intervals, they provide natural mobile sensing platforms at a fraction of the cost of dedicated survey vehicles.[2] The critical constraint introduced by this model is cost and robustness: sensors must be inexpensive enough for deployment across entire fleets, and must continue to function reliably in the same adverse conditions—heavy rain, fog, road dust, industrial smoke—during which road surfaces are most at risk.

It is precisely within this operational context that the CoSys-Lab research group at the University of Antwerp has investigated 3D in-air SONAR as a sensing modality. Their work, conducted under the imec.ICON Hybrid AI for Predictive Road Maintenance (HAIROAD) project, funded jointly by imec and the Flemish Agency for Innovation and Entrepreneurship (VLAIO, project no. HBC.2023.0170), presents a fully annotated dataset and systematic multi-model comparison for both road surface material classification and road damage type detection tasks.[Primary Paper; 32] The HAIROAD project itself runs from October 2023 through September 2025 and seeks to automate acquisition and interpretation of road condition indicators across one of the world's densest municipal road networks—Flanders, Belgium, where up to 30% of municipal budgets are devoted to mobility and road infrastructure, yet less than 5% of municipalities have formally adopted structured road condition assessment practices.[32]

Related Work and Sensing Modality Landscape

Camera-Based Road Damage Detection

Visual camera sensing remains the most mature and widely deployed technology for automated pavement condition assessment. Trinh, Anwar, and Mercelis applied road area extraction and contrastive learning to RGB images for road surface condition classification using a segmentation model to isolate road from surrounding pixels, improving classification across all damage categories.[3] Shim et al. employed a lightweight hierarchical auto-encoder network to detect the presence of surface damages from camera images, though their approach did not address damage type identification or overall surface quality assessment.[4]

The National Academies of Sciences, Engineering, and Medicine published a comprehensive 2024 synthesis report on AI Applications for Automatic Pavement Condition Evaluation, finding that 3D laser scanning data has become the predominant approach for automated collection due to its ability to gather depth information and its relative robustness to lighting conditions compared to 2D imaging.[3-NAS] Oklahoma State University's PaveVision3D system exemplifies this trajectory, enabling rapid collection of millimeter-level 3D pavement data across entire lane widths. Despite this maturity, all camera and LiDAR modalities share a fundamental vulnerability: optical degradation in adverse weather.

Acoustic and Vibration-Based Sensing

Li et al. demonstrated that tire noise alone can be used to classify pavement as being in normal condition, exhibiting crack damage, or showing pothole damage, achieving 88.4% overall accuracy using a random forest model combined with gradient boosting—at essentially zero additional hardware cost given that microphones are increasingly standard on commercial vehicles.[5] Dong and Li combined smartphone accelerometers with GPS data to identify surface distortions, patching, potholes, and rutting on moving vehicles using k-means clustering, reaching 84% average accuracy.[6]

A parallel line of research from the smart cities domain has directly explored acoustic-ultrasonic road sensing. In work published in Computational Urban Science, a wheel-rim-mounted module integrating a microphone with an ultrasonic depth sensor was coupled to Multi-Layer Perceptron, SVM, and Random Forest classifiers to distinguish smooth, slippery, grassy, and rough road conditions, demonstrating the feasibility of microphone-based road surface discrimination in smart city deployment scenarios.[19]

Kim et al. conducted a direct predecessor study using conventional (single-channel) ultrasonic sensors mounted at the front of vehicles, coupling short-time Fourier transform features to deep neural networks to identify road surface type across eight surface categories, achieving accuracies exceeding 95%.[9] This result is particularly informative as a benchmark: the 3D SONAR approach investigated by CoSys-Lab may be understood as a substantial multi-channel generalization of this prior ultrasonic work, potentially unlocking spatial resolution and directional information not available to single-element sensors.

Radar and Combined Modalities

Sattar, Li, and Chapman evaluated both 24 GHz RADAR and 40 kHz SONAR for classifying five road surface types, achieving up to 80% accuracy with individual modalities and 92% combined.[10] A 2025 study published in Sensors MDPI specifically examined automotive radar range spectra for road surface classification using Random Forest classifiers, achieving 84.5% generalization error under dry conditions and 88.7% distinguishing wet from dry asphalt, while also noting that no radar sensor is currently available commercially specifically for road classification purposes.[27] The 2025 IET Radar, Sonar and Navigation review by Bystrov provides a comprehensive survey of automotive microwave sensor approaches to this classification problem, noting the complementary characteristics of active radar and passive radiometry.[24]

In automotive ultrasonic object sensing—a related but distinct domain—a 2025 Journal of the Acoustical Society of America study by Eisele et al. demonstrated that replacing conventional single-element parking sensors with compact 2×2 MEMS transducer arrays and applying CNN classifiers substantially improved obstacle classification in automotive surround sensing, underscoring the broader applicability of small-aperture array ultrasonic sensing with machine learning.[14]

Drone- and UAV-Based Monitoring

A systematic literature review published through ASCE in 2025, surveying 60 articles from a pool of 619 publications between 2014 and 2022 on drone-based road condition monitoring (D-RCM), found consistent evidence of cost and time savings, safety improvements, and improved spatial coverage compared to ground-based methods.[8-ASCE] LiDAR-equipped drones have been operationally deployed by Alaska DOT, Caltrans, New York Thruway Authority, and Utah DOT for bridge and elevated corridor inspection with BVLOS (Beyond Visual Line of Sight) flights, reducing lane closures and enhancing inspector safety.[6-ALG]

While UAV-LiDAR represents a powerful high-resolution approach for targeted survey campaigns, it shares the limitation of all optically dependent modalities: performance degrades in fog, heavy precipitation, and smoke. It is also not well suited to the continuous, fleet-scale opportunistic data collection that is the principal motivation for the SONAR-based approach under review.

Sensor Technology and System Architecture

The eRTIS Platform

The eRTIS (Embedded Real-Time Imaging Sonar) is a fully embedded 32-channel 3D in-air SONAR sensor developed at the CoSys-Lab, University of Antwerp.[11,12,13,25,26] Inspired by bat echolocation, the system emits a hyperbolic broadband chirp spanning 20 kHz to 50 kHz, generated at 450 kHz sample rate with a 2.5 ms pulse duration, and records the resulting wavefield reflections across a 32-element MEMS microphone array at 4.5 MHz, producing 163,840 samples per capture. A dedicated on-board compute unit handles real-time signal processing, including PDM (Pulse Density Modulation) decoding, matched filtering, delay-and-sum beamforming across 91 directions spanning the full ±90° frontal hemisphere in both azimuth and elevation, envelope detection, and an artifact suppression stage analogous to a constant false-alarm rate (CFAR) detector. The processed output is a matrix of acoustic backscatter intensity as a function of angle and range, termed an energyscape.[Primary Paper]

The sensor architecture is modular: a high-performance microcontroller manages excitation signal generation and data acquisition, while an NVIDIA Jetson compute module handles GPU-accelerated beamforming and post-processing. An anodized aluminum enclosure with passive cooling and IP-rated connectors provides environmental protection suitable for vehicle mounting in harsh outdoor conditions.[26] The sensor is mounted on the rear of the vehicle, oriented downward and toward the road, with elevation angles sweeping from left to right from the driver's perspective and azimuth sweeping up and down—a geometry that captures the road surface immediately behind the vehicle.

Signal Processing Pipeline

Two parallel preprocessing pipelines are employed depending on the downstream classifier. For conventional machine learning models (logistic regression, SVM, random forest, gradient-boosted trees, multi-layer perceptron), a vector preprocessing pipeline applies max-pooling along the range dimension to the energyscape, flattens the resulting matrix, and applies whitened PCA reducing dimensionality to 256 principal components. For the CNN-based classifier, an image preprocessing pipeline introduces time-shift data augmentation (±45 samples at 450 kHz, corresponding to ±100 μs delays) to improve range-invariance, followed by per-energyscape normalization and random horizontal/vertical flips. Mean-subtraction across training samples is applied in both pipelines to emphasize inter-sample variation over absolute backscatter level.[Primary Paper]

Dataset Construction and Experimental Design

The dataset was collected using vehicle-mounted eRTIS sensors operating at approximately 10 Hz sample rate, simultaneously recording camera images used exclusively for labeling. Temporal synchronization was enforced with a 150 ms maximum allowed offset; samples exceeding this threshold were discarded. Labels encode both material type (asphalt, concrete, element) and up to nine damage categories (transversal crack, alligator crack, missing material, longitudinal crack, fraying, open longitudinal joint, subsidence, open transversal joint, loose stones), with multi-label capability to handle transitions between surface types and co-occurring damage types within a single acquisition.

Class filtering eliminated any damage category with fewer than 100 samples to reduce dataset imbalance. The full dataset was split into 10% held-out test set and 90% training/validation partitioned into 10 stratified folds with 90%/10% split per fold. The resulting test set contains 513 samples versus approximately 4,386 training samples per fold—a size asymmetry noted by the authors as a significant factor in the test-to-validation performance gap observed across models. Stratification was conducted on a synthetic integer derived from the one-hot encoded multi-label combination, thereby approximately preserving class distributions across splits.[Primary Paper]

A critical methodological limitation acknowledged by the authors is that dataset splitting does not respect the geographic origin of samples. Data points from physically proximate locations on the same road segment may be distributed across training, validation, and test sets—introducing potential data leakage and reducing the generalizability assessment to novel road environments. The authors recommend road-segment-based clustering as a prerequisite for future dataset construction.

Dataset Summary

• ~5,386 total samples; 513 held-out test, ~4,386 training per fold
• Labels: 3 material types × 9 damage categories (multi-label)
• 10-fold stratified cross-validation; each model trained twice with different random seeds
• Annotation source: synchronized camera images (not used for inference)
• Sampling rate: ~10 Hz; synchronization window: ±150 ms

Classifier Comparison and Results

Material Classification

Across logistic regression, decision tree, random forest, support vector machine (SVM), multi-layer perceptron (MLP), and gradient-boosted trees (GBT), all non-linear models substantially outperformed the linear baseline, confirming that the relationship between PCA components of acoustic energyscapes and road surface material type is inherently non-linear. The GBT model achieved the best test F1 score of 89.55% ± 0.60% and test κ of 78.10% ± 1.40%, consistent with pre-existing ultrasonic sensing literature showing above-90% accuracy for road type identification.[9, Primary Paper] Table I summarizes results.

Table I: Material Classification Performance (Averaged Across Folds and Random Initializations; μ ± σ)

Model	Test κ (%)	Val κ (%)	Test F1 (%)	Val F1 (%)
Logistic Regression	40.34 ± 1.22	48.00 ± 3.18	71.77 ± 0.90	75.73 ± 1.73
Decision Tree	76.82 ± 1.90	92.03 ± 0.48	89.03 ± 0.63	95.04 ± 0.47
Random Forest	76.84 ± 0.69	93.25 ± 1.46	89.44 ± 0.26	96.64 ± 0.54
SVM	74.67 ± 1.66	89.97 ± 2.59	88.03 ± 2.02	95.02 ± 1.73
MLP	77.52 ± 2.25	91.95 ± 2.21	89.12 ± 1.18	95.95 ± 1.08
Gradient-Boosted Trees	78.10 ± 1.40	93.58 ± 1.83	89.55 ± 0.60	96.79 ± 1.01

Damage Detection and Classification

The damage detection task proved substantially more challenging. GBT again led all models with test F1 of 71.74% ± 1.15% and test κ of 58.00% ± 2.38%. Notably, the CNN (ResNet-variant) model performed significantly worse than all other classifiers, with test F1 of only 28.68% ± 8.03%—a result attributed to a geometric discontinuity in the energyscape image representation arising from the interleaved azimuth/elevation indexing, which violates the spatial locality assumption underlying convolutional operations. Table II presents results.

Table II: Damage Detection Performance (Averaged Across Folds and Random Initializations; μ ± σ)

Model	Test κ (%)	Val κ (%)	Test F1 (%)	Val F1 (%)
Logistic Regression	25.88 ± 1.03	30.14 ± 2.35	49.16 ± 0.58	52.69 ± 1.17
Decision Tree	54.98 ± 1.51	65.23 ± 2.36	68.42 ± 0.78	75.76 ± 1.60
Random Forest	57.87 ± 1.46	65.96 ± 2.03	70.69 ± 0.69	76.40 ± 1.39
SVM	52.64 ± 1.25	59.90 ± 1.63	68.33 ± 1.01	73.15 ± 1.40
MLP	54.15 ± 1.82	63.54 ± 2.15	69.25 ± 0.95	75.83 ± 1.25
Gradient-Boosted Trees	58.00 ± 2.38	66.66 ± 2.66	71.74 ± 1.15	77.75 ± 1.54
ResNet (CNN)	2.57 ± 1.67	5.50 ± 2.18	28.68 ± 8.03	31.59 ± 8.35

Discussion and Critical Analysis

All-Weather Sensing Advantage

The fundamental value proposition of in-air SONAR for opportunistic road sensing lies not in raw classification accuracy relative to best-in-class optical systems under favorable conditions, but in its resilience under adverse conditions where optical systems fail. Fog, heavy rain, snow, smoke, and road dust—precisely the environmental conditions correlated with deteriorating road surfaces and increasing damage risk—cause minimal degradation to acoustic sensing. This characteristic directly enables the core operational requirement of an opportunistic sensing system: reliable, continuous data collection regardless of weather. Contemporary AI road monitoring reviews consistently identify optical degradation under adverse weather as a primary limitation constraining camera and LiDAR-based continuous monitoring.[6-ALG]

Performance Context and Benchmarking

The 90% F1 material classification result is directly in line with the Kim et al. (2021) prior result of over 95% accuracy using conventional ultrasonic sensing across eight road surface types,[9] and with the 92% combined RADAR+SONAR accuracy reported by Sattar et al.[10] These comparisons validate the approach and suggest that the 3D beamforming dimension of eRTIS—while adding spatial resolution and enabling concurrent obstacle detection—does not degrade acoustic surface characterization performance relative to simpler transducer configurations.

The 75% damage detection F1 is more difficult to contextualize because direct analogues are scarce in the literature. Camera-based damage detection typically reports higher accuracy on well-curated datasets but is sensitive to lighting, contrast, and viewing angle. The acoustic approach faces a different challenge: many surface damage types (fraying, loose stones, open joints) produce subtle backscatter differences that are at or below the detection threshold of a 10 Hz, vehicular-speed acquisition regime. The label distribution imbalance—several damage types appearing fewer than 100 times in the dataset—further constrains achievable performance on rarer classes.

CNN Architecture Failure Mode

The CNN (ResNet-variant) failure in damage detection is theoretically well-understood and practically important. The eRTIS energyscape is formed by concatenating rows from a 91-direction beamforming grid, but azimuth and elevation are collapsed into a single matrix dimension. Whenever the azimuth index wraps, a spatial discontinuity occurs in the 2D image representation. Convolutional kernels, operating under the assumption that adjacent pixels are spatially proximate, are therefore processing features across a geometric boundary—producing corrupted learned representations. Two remediation strategies are proposed: (1) reorganizing energyscapes into proper 3D tensors and applying 3D convolutions, at substantial computational cost; or (2) restricting beamforming to a 2D azimuthal slice, sacrificing some dimensional information but enabling valid 2D convolution. Neither approach was evaluated in the study, representing a clear priority for follow-on work.

Beamforming Algorithm Opportunities

The current implementation employs a linear delay-and-sum (DAS) beamformer—the simplest member of the beamforming family. More advanced algorithms—Minimum Variance Distortionless Response (MVDR) beamforming and Delay-Multiply-and-Sum (DMAS)—are known to produce substantially cleaner spatial spectra, reduced sidelobe contamination, and improved target contrast, at higher on-board computational cost.[24,25] Given the computational constraints of the eRTIS on-board processing unit, the authors identify advanced beamforming as a promising direction conditional on continued hardware capability improvements. The recent development of the HiRIS (High-Resolution Imaging Sonar) platform at CoSys-Lab—featuring a 1024-channel 32×32 uniform rectangular microphone array with MVDR forward-backward spatial smoothing—represents the architectural endpoint of this trajectory, offering 70 dB main lobe-to-sidelobe ratio in single-source scenarios.[25]

Dataset and Generalizability Constraints

Three structural dataset limitations warrant emphasis. First, annotation is based on camera images rather than certified ground-truth pavement survey data, introducing potential labeling bias toward visually obvious damage modes and under-representation of subsurface or low-contrast defects. Second, the absence of geographic stratification in dataset splitting means that physically correlated samples (from the same road segment) may leak across splits, potentially inflating reported performance. Third, the test set at 513 samples is small enough that a single misclassification in a rare class (which may appear only once or twice) can substantially shift the test κ for that class, creating high variance in the test performance estimate. Road-segment-level splitting and larger held-out geographic areas are essential prerequisites for credible generalizability claims.

Broader Policy and Deployment Context

The HAIROAD project is funded by imec and VLAIO (Flanders Innovation and Entrepreneurship) with a two-year execution window (October 2023–September 2025) and targets a Flemish road network context in which fewer than 5% of municipalities have formally adopted Belgian Road Research Centre (BRRC) standard road condition surveys, largely due to cost.[32] The sensor fusion and hybrid AI framework being developed through HAIROAD aims to automate BRRC indicator acquisition and supplement it with new indicators—including water management and debris accumulation—while providing forecasting capability sufficient to support cost-benefit-weighted maintenance recommendations.[32]

The broader European context features comparable initiatives. A 2025 AI blueprint review covering UK, US, and European highway agencies documented widespread adoption of AI vision systems on routine maintenance vehicles—exemplified by Hertfordshire County Council's 2024–2025 trials of the Robotiz3d ARRES Eye system, combining high-resolution cameras and LiDAR for real-time pothole, cracking, and rutting detection during routine van patrols.[39] In the United States, a 2024 National Academies report on AI applications for automatic pavement condition evaluation documented the accelerating transition from manual inspection to automated systems using laser profiling, accelerometer-based roughness measurement, and computer vision.[3-NAS] California's 2025 AI transportation initiative, announced by Governor Newsom, further signals political commitment to AI-driven infrastructure monitoring at scale.[39]

Within this landscape, SONAR-based opportunistic sensing occupies a distinctive niche: it complements rather than competes with these primarily optical approaches. Smart sensing platforms such as those developed by Vaisala—which aggregate accelerometer friction and roughness data from vehicle fleets in real time—and ASIMOB's accelerometer-derived impact index demonstrate that non-optical modalities are already penetrating mainstream PMS infrastructure at scale.[6-ALG] The CoSys-Lab eRTIS work extends this non-optical sensing trajectory into the acoustic imaging domain, with the potential to provide surface material discrimination and damage type classification beyond what accelerometers or single-element ultrasonic sensors currently deliver.

Future Research Directions

Recommended Research Priorities

• 3D Energyscape Representation: Restructure beamforming output into proper (azimuth × elevation × range) tensors and apply 3D CNNs to enable spatially valid convolution operations.
• Advanced Beamforming: Evaluate MVDR and DMAS beamformers within the on-board compute budget; assess impact on damage F1 across all damage categories.
• Geographic Dataset Stratification: Reprocess dataset with road-segment-level train/test splits to provide credible generalizability metrics on previously unseen road environments.
• Sensor Fusion: Combine eRTIS energyscapes with GPS, accelerometer, and camera inputs in a multi-modal fusion architecture to exploit complementary sensing strengths.
• Larger Balanced Dataset: Expand data collection to increase representation of rare damage classes (open transversal joint, subsidence, loose stones) and mitigate class imbalance effects on test performance stability.
• Damage Severity and Localization: Extend the classification framework to include severity grading and spatial localization of damage within the road surface footprint.
• Adverse Weather Validation: Formally characterize the all-weather performance advantage of SONAR versus optical sensors under quantified fog, rain, and dust conditions.

Conclusion

This review has examined the University of Antwerp CoSys-Lab contribution on 3D in-air SONAR sensing for opportunistic road condition monitoring within the broader context of pavement management systems research. The study establishes that the eRTIS 32-channel MEMS sonar array, combined with gradient-boosted tree classifiers operating on PCA-reduced energyscape features, achieves road material classification F1 scores approaching 90% and road damage detection F1 scores approaching 75%—performance levels that are promising and competitive with alternative acoustic modalities, while remaining below the thresholds required for industrial deployment without human verification.

The core technical value of the SONAR modality—imperviousness to optical degradation in adverse weather—provides a genuinely distinct capability for opportunistic sensing deployments compared to camera, LiDAR, or hybrid visual systems. Identified paths to improvement include advanced beamforming algorithms, geometrically valid 3D convolutional architectures, road-segment-stratified dataset construction, and multi-modal sensor fusion. Within the HAIROAD project framework and the broader European commitment to AI-driven infrastructure management, this work represents a meaningful and technically rigorous step toward cost-effective, continuous, all-weather pavement management at municipal scale.

The primary work reviewed herein was supported by the imec.ICON Hybrid AI for Predictive Road Maintenance (HAIROAD) project, funded by imec and Flanders Innovation & Entrepreneurship (VLAIO), project no. HBC.2023.0170. All authors of the primary paper are affiliated with the Department of Electronics and ICT Engineering Technology, CoSys-Lab Research Group, University of Antwerp, Antwerp, Belgium, and Flanders Make Strategic Research Centre, Lommel, Belgium.

---

References

[1] H. Shon, C.-S. Cho, Y.-J. Byon, and J. Lee, "Autonomous condition monitoring-based pavement management system," Automation in Construction, vol. 138, p. 104 222, 2022. DOI: 10.1016/j.autcon.2022.104222. https://www.sciencedirect.com/science/article/pii/S0926580522000954

[2] W. Hu, S. Winter, and K. Khoshelham, "Forecasting fine-grained sensing coverage in opportunistic vehicular sensing," Computers, Environment and Urban Systems, vol. 100, p. 101 939, 2023. DOI: 10.1016/j.compenvurbsys.2023.101939. https://www.sciencedirect.com/science/article/pii/S0198971523000029

[3] L. Trinh, A. Anwar, and S. Mercelis, "Improving classification of road surface conditions via road area extraction and contrastive learning," IECON 2024 – 50th Annual Conference of the IEEE Industrial Electronics Society, 2024, pp. 1–7. DOI: 10.1109/IECON55916.2024.10905278.

[4] S. Shim, J. Kim, S.-W. Lee, and G.-C. Cho, "Road surface damage detection based on hierarchical architecture using lightweight auto-encoder network," Automation in Construction, vol. 130, p. 103 833, 2021. DOI: 10.1016/j.autcon.2021.103833. https://www.sciencedirect.com/science/article/pii/S0926580521002843

[5] H. Li, R. Chen, N. Ritha, and J. Wang, "Research on intelligent detection system and method for road surface damage utilizing tire noise," Measurement, vol. 253, p. 117 650, 2025. DOI: 10.1016/j.measurement.2025.117650. https://www.sciencedirect.com/science/article/pii/S0263224125010097

[6] D. Dong and Z. Li, "Smartphone sensing of road surface condition and defect detection," Sensors, vol. 21, no. 16, 2021. DOI: 10.3390/s21165433. https://www.mdpi.com/1424-8220/21/16/5433

[7] D. Kim, R. Kim, J. Min, and H. Choi, "Initial freeze–thaw damage detection in concrete using two-dimensional non-contact ultrasonic sensors," Construction and Building Materials, vol. 364, p. 129 854, 2023. DOI: 10.1016/j.conbuildmat.2022.129854.

[8] R. Shi et al., "CNN-transformer for visual-tactile fusion applied in road recognition of autonomous vehicles," Pattern Recognition Letters, vol. 166, pp. 200–208, 2023. DOI: 10.1016/j.patrec.2022.11.023.

[9] M.-H. Kim, J. Park, and S. Choi, "Road type identification ahead of the tire using d-cnn and reflected ultrasonic signals," International Journal of Automotive Technology, vol. 22, no. 1, pp. 47–54, 2021. DOI: 10.1007/s12239-021-0006-6. https://link.springer.com/article/10.1007/s12239-021-0006-6

[10] S. Sattar, S. Li, and M. Chapman, "Developing a near real-time road surface anomaly detection approach for road surface monitoring," Measurement, vol. 185, p. 109 990, 2021. DOI: 10.1016/j.measurement.2021.109990. https://www.sciencedirect.com/science/article/pii/S0263224121009192

[11] D. Laurijssen, W. Jansen, A. Aerts, and J. Steckel, "Ruggedized ultrasound sensing in harsh conditions: eRTIS in the Wild," IEEE Sensors Journal, Jan. 2025. arXiv: 2509.10029. https://arxiv.org/abs/2509.10029

[12] R. Kerstens, D. Laurijssen, and J. Steckel, "Ertis: A fully embedded real time 3d imaging sonar sensor for robotic applications," in 2019 International Conference on Robotics and Automation (ICRA), pp. 1438–1443. DOI: 10.1109/ICRA.2019.8794419. https://www.researchgate.net/publication/335140721

[13] J. Steckel, A. Boen, and H. Peremans, "A sonar system using a sparse broadband 3d array for robotic applications," in 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 3223–3228. DOI: 10.1109/IROS.2012.6385584.

[14] J. Eisele, A. Gerlach, Y. Manz, M. Maeder, and S. Marburg, "Object classification in automotive ultrasonic surround sensing using a compact 2×2 sensor array and a convolutional neural network," Journal of the Acoustical Society of America, vol. 157, no. 4, pp. 2556–2569, Apr. 2025. DOI: 10.1121/10.0036385. https://pubs.aip.org/asa/jasa/article/157/4/2556/3342776

[15] T. Chen and C. Guestrin, "XGBoost: A scalable tree boosting system," in Proc. 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 785–794. DOI: 10.1145/2939672.2939785. http://doi.acm.org/10.1145/2939672.2939785

[16] A. Paszke et al., "Automatic differentiation in pytorch," in NIPS-W, 2017.

[17] C. Ying, A. Klein, E. Christiansen, E. Real, K. Murphy, and F. Hutter, "NAS-bench-101: Towards reproducible neural architecture search," in Proc. 36th International Conference on Machine Learning, PMLR, vol. 97, 2019, pp. 7105–7114. https://proceedings.mlr.press/v97/ying19a.html

[18] M. Tan and Q. Le, "EfficientNet: Rethinking model scaling for convolutional neural networks," in Proc. 36th International Conference on Machine Learning, PMLR, vol. 97, 2019, pp. 6105–6114. https://proceedings.mlr.press/v97/tan19a.html

[19] Authors, "Artificial intelligence for road quality assessment in smart cities: a machine learning approach to acoustic data analysis," Computational Urban Science, 2023. DOI: 10.1007/s43762-023-00104-y. https://link.springer.com/article/10.1007/s43762-023-00104-y

[20] J. Steckel, A. Boen, and H. Peremans, "Broadband 3-d sonar system using a sparse array for indoor navigation," IEEE Transactions on Robotics, vol. 29, no. 1, pp. 161–171, 2013. DOI: 10.1109/TRO.2012.2221313.

[24] A. Bystrov et al., "Microwave sensor technologies for road surface classification: A comprehensive review," IET Radar, Sonar & Navigation, 2025. DOI: 10.1049/rsn2.70080. https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.70080

[25] CoSys-Lab, University of Antwerp, "Our Sensors: eRTIS and HiRIS," RTIS Sensor Platform Technical Documentation, accessed March 2026. https://www.3dsonar.eu/copy-of-applications

[26] D. Laurijssen, W. Jansen, A. Aerts, and J. Steckel, "eRTIS: A Fully Embedded Real Time 3D Imaging Sonar Sensor for Robotic Applications," ResearchGate, 2019. https://www.researchgate.net/publication/335140721

[27] Authors, "Automotive radar range spectrum-based road surface classification by using machine learning," Sensors, vol. 25, no. 22, p. 6911, Nov. 2025. DOI: 10.3390/s25226911. https://www.mdpi.com/1424-8220/25/22/6911

[32] imec, "HAIRoad – imec.icon project: Hybrid AI for Predictive Road Maintenance," imec Research Portfolio, 2023–2025. https://www.imec-int.com/en/research-portfolio/hairoad

[NAS] National Academies of Sciences, Engineering, and Medicine, "AI Applications for Automatic Pavement Condition Evaluation," Washington, DC: The National Academies Press, 2024. DOI: 10.17226/27993. https://www.nationalacademies.org/read/27993/chapter/4

[ALG] ALG Global, "Road monitoring in the age of AI: challenges, solutions, and regional perspectives," ALG Technical Blog, 2025. https://www.alg-global.com/blog/land/road-monitoring-age-ai-challenges-solutions-and-regional-perspectives

[39] Highways Today, "AI's Blueprint for the Smart Highway Revolution," May 2025. https://highways.today/2025/05/28/ai-blueprint-for-smart-highway/

[ASCE] ASCE, "Evaluating the potential of drones to monitor road conditions," Journal of Transportation Engineering, Part B: Pavements, Mar. 2025. DOI: 10.1061/JPEODX.PVENG-1559. https://www.asce.org/

[Primary] A. Cassimon, R. Kerstens, W. Daems, and J. Steckel, "Intelligent Road Condition Monitoring using 3D In-Air SONAR Sensing," arXiv:2603.28141v1 [cs.CV], Mar. 30, 2026. https://arxiv.org/abs/2603.28141