In connected car news are Pony.ai, GAC, Tesla, Waymo, Ouster, Lucid and Bolt.
Pony.ai and GAC Gen-4 Robotruck
At IAA Transportation 2026 in Hannover, Pony.ai and GAC Commercial Vehicle pulled the cover off a Level 4 autonomous heavy-duty truck that signals how quickly self-driving freight is moving from demonstration to deployment. Built on GAC’s T9 battery-electric heavy-duty platform, the truck marries Pony.ai’s Gen-4 autonomous system to a drive-by-wire chassis engineered specifically for driverless operation, with volume manufacturing slated to begin later this year.
The companies are targeting long-haul corridors, dedicated logistics routes, and port drayage — the kind of repetitive, high-mileage routes that have long been considered the low-hanging fruit of autonomous trucking. Underpinning the system is a fully redundant architecture spanning steering, braking, communications, power distribution, compute, and sensors, backed by a perception suite of nine lidars, three millimeter-wave radars, and 13 cameras for 360-degree coverage.
Pony.ai is leaning on the same domain controller architecture used in its Gen-7 Robotaxis, a decision the company says cuts the autonomous driving kit’s bill of materials by 70 percent compared with its prior generation. Combined with a 0.4 drag coefficient body and proprietary energy-dispatch algorithms that trim fleet power consumption by 10 percent, the companies project a 30 percent reduction in ton-kilometer transportation costs. The truck also marks the first extension of Pony.ai’s vehicle-agnostic Virtual Driver platform into European and Middle Eastern logistics markets, with rollout planned over the next two years.
Tesla Faces New Pressure Over Redacted Crash Data
Not every corner of the autonomous vehicle world is moving in lockstep toward transparency. Trial attorney Amy Witherite, founder of Witherite Law Group, has publicly challenged Tesla’s practice of redacting crash telematics submitted to the National Highway Traffic Safety Administration, reigniting a debate over how much automakers should be allowed to shield from public view.
The dispute traces back to a fatal crash on October 31, 2025, when a 2020 Tesla Model 3 came to a stop in an active travel lane on Arizona’s Loop 202 with its automated driving system verified as engaged. A Ford F-350 struck the stationary Tesla from behind, killing the Tesla’s driver. In its regulatory filings, Tesla classified core operational details — the crash narrative, the active software version, and the system’s operational design domain parameters — as confidential business information.
Witherite argues that proprietary protections should not be allowed to obscure the kind of data that determines what actually happened in a crash: sensor detection logs, warning activations, and the timeline of any system disengagement. She frames the redactions as part of a broader information imbalance, in which manufacturers control the telematics, video, and event data recorder logs that crash victims and their families need to understand what went wrong.
The scrutiny is not limited to one law firm. NHTSA opened a formal audit in August 2025 into whether Tesla’s automated driving system crash reports meet its Standing General Order requirements for timeliness and completeness. In a March 2026 filing covering eight separate incidents, Tesla asked for perpetual confidential treatment of its detailed reporting — a request that has drawn renewed criticism from traffic safety advocates pushing for greater fleet-wide transparency.
Waymo Sets Its Sights on Tokyo
Waymo is preparing for its first fully driverless commercial launch in Asia. The Alphabet subsidiary has finalized an agreement with Japanese taxi operator Nihon Kotsu and mobility platform GO to bring Level 4 robotaxi service to Tokyo by 2027, building on supervised mapping and road testing that began in the city in 2025.
That early testing relied on Nihon Kotsu safety operators to help train the Waymo Driver on Tokyo’s narrow residential streets and dense pedestrian environments — terrain quite different from the wide boulevards of Phoenix or San Francisco. Once launched, rides will be dispatchable through both the GO app and Waymo’s own app, with an initial commercial fleet expected to scale to roughly 100 vehicles spread across major Tokyo wards. Nihon Kotsu will handle depot infrastructure, maintenance, and ground logistics, a division of labor designed in part to offset labor shortages tied to Japan’s aging population.
The launch still hinges on permits from Japanese regulatory ministries and regional transport authorities, but it represents a milestone for Waymo, extending its driverless footprint beyond 15 U.S. metropolitan markets and into its first international commercial deployment.
Ouster’s New Lidar Brings Color to the Point Cloud
On the hardware side of the industry, digital lidar maker Ouster has introduced the Rev8 OS1 Max, a 256-channel sensor that integrates native 48-bit color point acquisition directly onto its silicon. Built around the company’s proprietary L4 Max system-on-chip, the sensor performs spatial-temporal fusion at the hardware level, registering RGB color data with 3D range coordinates point by point — eliminating the multi-sensor calibration headaches, parallax errors, and texturing lag that have complicated mobile mapping workflows.
The numbers are striking: 1-sigma range precision of plus-or-minus 0.25 centimeters, absolute accuracy of plus-or-minus 1.25 centimeters, and a dynamic range of 116 decibels that lets the sensor operate from near-total darkness to 2 million lux. It can output up to 10.4 million points per second and detect objects out to 200 meters at 10 percent surface reflectivity, or up to 500 meters under ideal conditions — roughly twice the precision of Ouster’s previous Rev7 generation.
Ouster is already putting the sensor to work through a partnership with geospatial payload provider GeoCue, integrating the Rev8 OS1 Max into GeoCue’s TrueView drone hardware and LP360 processing software. Testing in Huntsville, Alabama, showed the combination could resolve thin utility conductors, structural edges, and power lines during high-altitude corridor flights, all while meeting Build America, Buy America Act and NDAA Section 164 domestic sourcing requirements.
Lucid and Bolt Bet Partner for European Robotaxis
Lucid Group is taking its software-defined vehicle ambitions to Europe through a new partnership with ride-hailing platform Bolt. The two companies plan to co-develop and deploy at least 25,000 SAE Level 4 autonomous vehicles across European cities, built on Lucid’s forthcoming Midsize platform — a piece of Bolt’s larger goal of operating 100,000 autonomous vehicles by 2035.
The vehicles will be built around NVIDIA’s Hyperion reference architecture, with centralized compute and a standardized sensor suite designed into the platform from the outset rather than retrofitted later. A newly formed Lucid Technologies division will work alongside Bolt Autonomous Driving Solutions to align hardware redundancies, safety systems, and telematics with European regulatory requirements.
Bolt will own and operate the resulting fleet, drawing on operational data from the 850 cities where it already runs ride-hailing service to shape software parameters, passenger interfaces, and dispatch systems. Both companies say they intend to work directly with regional policymakers to establish compliance pathways as commercial autonomous service scales across major European transit networks.
Bias Found in Self-Driving Systems See Pedestrians
A study from researchers at King’s College London is raising uncomfortable questions about equity in autonomous vehicle perception. The research, which examined the large language models and vision-language models increasingly used in AV decision-making, found measurable variance in pedestrian yielding rates — with braking decisions correlating to demographic markers, skin tone, and disability status rather than to kinematic factors like speed and distance alone.
The findings, covered by Automotive News and CarBuzz, echo concerns long raised by disability advocates. The Disability Rights Education & Defense Fund has noted that while today’s autonomous driving stacks are generally reliable at identifying standing or walking pedestrians, their object classification networks often struggle with individuals using wheelchairs, mobility scooters, or white guide canes — precisely the edge cases where safe, predictable behavior matters most.
In response, the King’s College London research team has proposed testing methodologies that strip demographic signifiers from model input layers, an approach aimed at ensuring that path-planning decisions are made on safety-critical factors rather than on who or what an AI system perceives a pedestrian to be.