InfrastructureIssue 136 records
Transportation
Six transportation AI signals not previously archived. NHTSA opened a recall query (Nov. 6 filing, reported Nov. 9) after most Waymo school-bus passing incidents occurred following the software fix. In Singapore, LTA also paused Grab's junction testing after Nov. 5 near misses during new on-demand route trials; residents worried but backed continued trials. Bangor, Maine, runs 42 Miovision AI detection cameras at signals at 400P resolution to rule out facial ID and plate reading. Lincoln County, S.D., approved a $34,000, three-year AI road-weather camera contract on Nov. 10 despite resident surveillance concerns and a commissioner's no-savings point. Goyang, Korea, starts a five-intersection real-time AI signal pilot Nov. 23, and Korea's TS showcased AI risk detection for unmanned Yangsan light rail stations opening Dec. 30. No measured results; no tolling, parking, port or fleet items qualified today.
- Evidence records
- 6
- Cross-source patterns
- 2
- Evidence classes
- 6 independent reporting
- Outcomes
- 3 emerging2 cautionary1 mixed
- Source freshness
- 6 new this fortnight
- Research completed
- 2026-11-11
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
AV safety fixes and service expansions are being questioned after the fact
NHTSA opened a recall query because most reported Waymo school-bus passing incidents came after a software fix was declared complete. In Punggol, both Nov. 5 near misses happened while AVs tested new on-demand routes, leading LTA to impose targeted pauses on Grab and ComfortDelGro. In both cases regulators acted on incident evidence after operators had changed software or service scope, not on pre-change validation.
Operating questionWhat evidence must an AV operator provide, and who checks recurrence, before a remedy is accepted or a new route or maneuver is allowed?
Supporting evidenceNational Highway Traffic Safety Administration (reported by TechCrunch)Land Transport Authority, Singapore (reported by CNA)
Local roadside AI cameras are meeting surveillance questions, answered with design choices rather than published policy
Bangor said its signal detection cameras deliberately use 400P resolution so facial ID and plate reading are impossible, after residents asked whether they were Flock cameras. In Lincoln County, residents questioned AI road-weather cameras and the superintendent did not rule out law-enforcement access before the contract was approved. Neither source shows a written data-use or retention policy.
Operating questionBefore installing any roadside AI camera, has the agency published what it captures, who may access it and how long it is kept?
Supporting evidenceCity of Bangor, MaineLincoln County Commission and Highway Department, South Dakota
Evidence ledger
Full record · every source keeps its link and limitations
NHTSA opens a recall query into whether Waymo's school-bus software fix worked, saying most incidents came after the update
TechCrunch reports that NHTSA's Office of Defects Investigation opened a recall query in a November 6 filing to assess the adequacy of Waymo's remedy for robotaxis illegally passing stopped school buses. Waymo said the software update was completed on all affected vehicles on November 17, 2025, but NHTSA said the majority of reported incidents occurred after that update. The original probe began in October 2025 after an Atlanta incident; Austin ISD reported at least five incidents. Waymo says it has received no citations since March 2026.
Why it matters, evidence and limitations
- Why it matters
- Directly relevant to state DOTs, school districts and cities that host or permit AV operations: a software recall was declared complete yet the unsafe behavior continued, so local agencies cannot treat a vendor's fix notice as verification. Federal oversight is the lever here; state and local roles are incident reporting and permit conditions.
- Evidence and measured results
- Secondary reporting of a federal filing. The article does not give total incident counts after the fix, the filing text or methodology; the NHTSA document itself was not opened. Waymo's no-citations claim is an operator statement.
- Limitations and uncertainty
- Secondary source; the NHTSA recall query document was not opened and the vehicle count reported elsewhere was not confirmed from the cited page. Investigation is open; no finding of defect adequacy yet.
Bangor, Maine, runs 42 Miovision AI detection cameras at signals and limits resolution to rule out facial ID and plate reading
Bangor has installed Miovision cameras that detect waiting vehicles so signals serve those movements, replacing aging detection that needed constant repairs. The system includes an AI engine, Mateo, that collects traffic data for engineers. After public questions, the city electrician said the cameras are not Flock cameras and deliberately use 400P resolution so facial identification and license plate reading are not possible. Forty-two cameras are active; installation took over three years and about a dozen more are planned within a year.
Why it matters, evidence and limitations
- Why it matters
- A small-city example of AI-enabled signal detection where public trust hinged on privacy design. Transferable to U.S. cities and counties replacing loop or older video detection, especially where ALPR controversies make residents suspicious of any roadside camera.
- Evidence and measured results
- Local news interview with a city official. The only performance statement is qualitative ('solved a lot of the issues'); no delay, crash or maintenance metrics, and no vendor documentation reviewed.
- Limitations and uncertainty
- Single local news source quoting one official; resolution claim not independently verified; no outcome data.
Lincoln County, S.D., approves $34,000 AI road-weather cameras over resident surveillance worries and a no-savings admission
On Nov. 10 the Lincoln County Commission approved a three-year, $34,000 contract with Frost Solutions for weather-station cameras at four intersections. The stations measure air and road temperature, dew point and humidity, take photos every 20 minutes, and use AI to detect snowfall, build local forecasts and answer highway workers' questions. The superintendent said they are not surveillance cameras but did not rule out sharing access with law enforcement. A commissioner noted the contract would be a wash on overtime, not a cost saving; another backed it on safety grounds.
Why it matters, evidence and limitations
- Why it matters
- Practical, low-cost AI for county road-weather decisions, with a candid record that the expected benefit is safety and convenience rather than savings. Transferable to rural counties and state DOT maintenance districts, where similar stations are already used in Sioux Falls and Minnehaha County.
- Evidence and measured results
- Local news account of a public meeting, including officials' statements and resident testimony. No measured outcomes; the no-savings point is a commissioner's reading of the superintendent's remarks.
- Limitations and uncertainty
- Single local news source; vote count and final contract terms beyond price and term were not detailed in the extracted text; vendor capabilities as described by the county.
Goyang, Korea, will pilot real-time AI signal timing at five Deogeun intersections from Nov. 23
Goyang said on Nov. 9 it will begin trial operation of a real-time AI traffic signal system at five intersections on Eutteum-ro in the Deogeun district from Nov. 23. Traffic volumes collected by direction-specific CCTV at smart intersections are analyzed by AI, which calculates optimal signal times in real time and applies them automatically in the field. No targets, baselines or evaluation plan are stated.
Why it matters, evidence and limitations
- Why it matters
- A small, scoped adaptive-signal pilot using existing smart-intersection CCTV. Useful for U.S. cities considering narrow corridor pilots before broader adaptive control; Korean procurement and camera norms differ.
- Evidence and measured results
- Brief news-agency report of a city announcement. No performance claims or methods; pilot has not started.
- Limitations and uncertainty
- Korean-language source read in original; brief article; pilot not yet started; no evaluation design reported.
Korea's transport safety authority showcases AI risk detection for unmanned stations on the Yangsan light rail before its Dec. 30 opening
At a Nov. 9–10 TS workshop, Busan Transportation Corporation demonstrated AI video analysis, an integrated operation management system and unmanned security systems for the Yangsan Urban Railway, an 11.43 km unmanned light rail with seven stations opening Dec. 30. Demonstrations covered support for mobility-impaired riders, real-time recognition of risky passenger behavior and incident response procedures. Revenue service tests ran Sept. 5–Oct. 3, and comprehensive test results are still being verified. No AI performance data were reported.
Why it matters, evidence and limitations
- Why it matters
- Relevant to U.S. transit agencies weighing reduced station staffing or automated lines: AI video monitoring is being positioned as the safety and accessibility layer for unmanned stations. Korean labor, privacy and operating rules differ, limiting direct transfer.
- Evidence and measured results
- News report of a government workshop and demonstrations. No accuracy, response time or incident data; test results under verification.
- Limitations and uncertainty
- Article machine-translated by the publisher; single source; demonstrations, not evaluated results; non-U.S. context.
After two Punggol AV near misses, LTA also paused Grab's junction testing, and residents voiced concern but backed continued trials
CNA reports two Nov. 5 near misses in Punggol: a Grab AV came close to a girl crossing during a discretionary left turn, and a ComfortDelGro AV turned into an oncoming vehicle while attempting a U-turn. Both vehicles were testing new routes for on-demand services, not their fixed shuttle routes. LTA is investigating, directed Grab to pause testing at such junctions and told ComfortDelGro to pause U-turns at the junction. Fifteen passenger AVs operated in Punggol as of Nov. 3. Of 12 residents interviewed, parents were most concerned, but most supported trials continuing.
Why it matters, evidence and limitations
- Why it matters
- Extends the archived LTA ComfortDelGro pause with the Grab pause and resident sentiment. Relevant to U.S. transit agencies piloting AV shuttles: incidents clustered when vehicles moved off fixed routes into new on-demand paths, and targeted pauses preserved public support.
- Evidence and measured results
- News reporting with resident interviews (12 people, not a representative sample) and official actions. No investigation findings yet.
- Limitations and uncertainty
- Overlaps an archived item on the ComfortDelGro pause; new content is the Grab pause, operating context and resident views. Interview sample is small and self-selected; non-U.S. context.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.