{"id":"fd274a52-c3a2-42f7-a99d-8b78abfdd92a","slug":"nhtsa-sgo-30413-13249","kind":"collision","headline":"ADS crash report in Atlanta, GA involving a TOYOTA Sienna (NHTSA SGO 30413-13249)","summary":"During normal autonomous operation, the May autonomous shuttle (Ego) was traveling through a narrow section of roadway in a parking lot bordered by parked vehicles on the left and a retaining wall on the right on experimental software version v29.0.20. As the Ego proceeded, the system initially recognized the static retaining wall as a large, non-rectangular object. As the Ego approaches the wall, the resulting density of lidar returns on the object (wall) exceeds the detectors maximum limitation. The newer lidar points, toward the forward edge of the elongated object, exceeding the limit were filtered out, causing the track to disappear when the ego approached 5m from the track (i.e. the wall). The situation was exacerbated by the fact that the road edge was incorrectly marked in that section.  The road edge was drawn through the retaining wall and thus the vehicle did not have any prior knowledge about the edge of the road. As a result, the Ego drove too close to the wall, scraping it at a speed of 2 meters per second while driving straight (i.e. in lane-nominal policy) with no nudge or veer corrections. There was no damage to the wall. The limitation leading to the lost track affects Software versions v26.0 (when the Open World Tracklet v2 feature was introduced) through v31.0. This is the only known instance of this failure mode, and in v32.0 an optimization was made to remove the LIDAR threshold entirely and mitigate the vulnerability.  Furthermore, the incorrect road edge annotation was corrected shortly after this incident and the map updated for the deployed vehicles.","occurredAt":"2025-11-01T12:00:00.000Z","reviewStatus":"reviewed","modelId":null,"deploymentId":null,"companyId":"8f52c3ed-fb8c-4f25-8c75-c7c9f912776d","aliases":[],"collisionRisk":"none","reviewNote":null,"severity":"minor","rootCauseCategory":null,"rootCauseDetail":null,"affectedUnitsCount":null,"affectedUnitsBasis":null,"remediationStatus":null,"regulatoryAction":null,"regulatoryBody":null,"propertyDamageEstimateUsd":null,"damageEstimateBasis":null,"bodilyInjurySeverity":"unknown","fatalityCount":null,"lossCostClass":null,"sourceQualityTier":null,"oshaRecordable":null,"oshaCaseNumber":null,"rootCauseDisclosed":null,"publicDisclosureDate":null,"insuranceClaimPublic":null,"createdAt":"2026-07-12T05:02:00.031Z","updatedAt":"2026-07-14T21:53:38.800Z","model":null,"deployment":null,"company":{"id":"8f52c3ed-fb8c-4f25-8c75-c7c9f912776d","name":"May Mobility","slug":"may-mobility","description":"May Mobility is an autonomous-shuttle / microtransit company running driverless public AV services (autonomous Toyota Sienna minivans). Founded in 2017 and anchored by strategic backer Toyota.","status":"Private","founded":2017,"hq":"Ann Arbor, Michigan, USA","fundingTotal":null,"type":null,"reviewStatus":"reviewed","lifecycleState":"active","supersededByCompanyId":null,"sources":[{"url":"https://maymobility.com/","title":"May Mobility","sourceName":"May Mobility (official)"},{"url":"https://en.wikipedia.org/wiki/May_Mobility","title":"May Mobility","sourceName":"Wikipedia"}],"keyFacts":[{"label":"Founded","value":"2017 by Edwin Olson, Steve Vozar and Alisyn Malek"},{"label":"Backers","value":"Toyota (led Series B; Toyota AI Ventures), BMW i Ventures, SPARX, Millennium, Cyrus Capital"},{"label":"Deployments","value":"Driverless public service Peachtree Corners GA (2025); rider-only Sun City AZ (2023); earlier Detroit/Columbus/Providence/Grand Rapids"},{"label":"Vehicle platform","value":"Autonomous Toyota Sienna minivans"},{"label":"Business type","value":"Autonomous-shuttle / microtransit company"},{"label":"Service type","value":"Driverless public AV services"}],"aliases":["May Mobility"],"collisionRisk":"low","reviewNote":null,"atsProvider":"greenhouse","atsSlug":"maymobility","factoryLocations":null,"manufacturingCapacity":null,"orgType":null,"burnSignal":null,"litigationEvents":null,"insurancePostureDisclosed":null,"unitEconomicsDisclosed":null,"createdAt":"2026-06-05T03:46:40.621Z","updatedAt":"2026-06-29T02:33:17.662Z"},"derived":{"status":"active"},"sources":[{"id":"a3f06ba8-b7b1-4c1d-90b1-cff0c4e88e04","incidentId":"fd274a52-c3a2-42f7-a99d-8b78abfdd92a","url":"https://www.nhtsa.gov/laws-regulations/standing-general-order-crash-reporting#report-30413-13249","label":null,"publishedAt":null,"notes":null,"title":null,"sourceName":null,"mediaType":"article","videoId":null,"channel":null,"durationSec":null,"timestampSec":null,"createdAt":"2026-07-12T05:02:00.057Z","updatedAt":"2026-07-12T05:02:00.057Z"}],"responses":[],"statusEvents":[],"jsonLd":{"@context":"https://schema.org","@type":"NewsArticle","@id":"https://registry.deploy.report/incidents/nhtsa-sgo-30413-13249","url":"https://registry.deploy.report/incidents/nhtsa-sgo-30413-13249","headline":"ADS crash report in Atlanta, GA involving a TOYOTA Sienna (NHTSA SGO 30413-13249)","description":"During normal autonomous operation, the May autonomous shuttle (Ego) was traveling through a narrow section of roadway in a parking lot bordered by parked vehicles on the left and a retaining wall on the right on experimental software version v29.0.20. As the Ego proceeded, the system initially recognized the static retaining wall as a large, non-rectangular object. As the Ego approaches the wall, the resulting density of lidar returns on the object (wall) exceeds the detectors maximum limitation. The newer lidar points, toward the forward edge of the elongated object, exceeding the limit were filtered out, causing the track to disappear when the ego approached 5m from the track (i.e. the wall). The situation was exacerbated by the fact that the road edge was incorrectly marked in that section.  The road edge was drawn through the retaining wall and thus the vehicle did not have any prior knowledge about the edge of the road. As a result, the Ego drove too close to the wall, scraping it at a speed of 2 meters per second while driving straight (i.e. in lane-nominal policy) with no nudge or veer corrections. There was no damage to the wall. The limitation leading to the lost track affects Software versions v26.0 (when the Open World Tracklet v2 feature was introduced) through v31.0. This is the only known instance of this failure mode, and in v32.0 an optimization was made to remove the LIDAR threshold entirely and mitigate the vulnerability.  Furthermore, the incorrect road edge annotation was corrected shortly after this incident and the map updated for the deployed vehicles.","identifier":"fd274a52-c3a2-42f7-a99d-8b78abfdd92a","datePublished":"2025-11-01T12:00:00.000Z","articleSection":"collision","publisher":{"@id":"https://deploy.report/#organization"},"author":{"@id":"https://deploy.report/#organization"}},"framework_metadata":{"framework_schema_version":"0.1.0","verification_status":"verified","maturity_stage":null,"lifecycle_state":null,"architectural_position":{"cohort":null,"sub_cohorts":[]},"within_cohort_verified_vs_claimed_pair":null,"cap_flags":[],"verification_depth":{"sources_count":1,"primary_source_types":["primary-regulatory-filing"]}}}