AI in transportation
Last reviewed
Sources
No citations yet
Review status
Needs citations
Revision
v2 · 4,471 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
No citations yet
Review status
Needs citations
Revision
v2 · 4,471 words
Add missing citations, update stale details, or suggest a clearer explanation.
AI in transportation refers to the application of artificial intelligence techniques to the movement of people and goods across road, rail, air, sea, and the spaces in between. It spans far more than self-driving cars. Modern transportation systems use machine learning, computer vision, optimization, and forecasting to time traffic signals, route delivery fleets, predict travel demand, schedule maintenance on locomotives and aircraft, plan fuel-efficient flight paths, avoid collisions at sea, and operate sidewalk delivery robots.
The field draws on several decades of work in operations research and intelligent transportation systems (ITS), a term used since the 1990s for the integration of sensing, communications, and computation into surface transport. The deep learning era that began in the 2010s broadened what was possible, enabling perception systems that interpret raw camera and sensor data and large statistical models that forecast demand and travel times at the scale of entire cities. As of 2026, AI is embedded in everyday transportation infrastructure that most travelers never see, from the routing algorithm behind a parcel delivery to the model that estimates a ride-hailing arrival time.
Autonomous road vehicles are one prominent application but only a fraction of the whole. This article surveys AI across transportation modes and treats self-driving cars in summary, cross-linking the dedicated article on autonomous driving rather than repeating it.
The use of computation to manage traffic predates the modern AI boom. Adaptive traffic-signal control systems emerged in the 1970s, most notably the Sydney Coordinated Adaptive Traffic System (SCATS), developed in Sydney, Australia, and the SCOOT (Split Cycle Offset Optimisation Technique) system developed by the United Kingdom's Transport Research Laboratory. These systems adjusted signal timing in response to measured traffic flow rather than running fixed schedules, and they spread internationally during the 1980s and 1990s. SCATS became operational in Troy, Michigan in June 1992, an early United States deployment.1
In the United States, the field of intelligent transportation systems was formalized through the Intermodal Surface Transportation Efficiency Act of 1991, which funded research into applying information technology to highways. The wider availability of civilian GPS during the 1990s, combined with falling sensor and computing costs, made it feasible to collect and act on real-time traffic data at scale.2
Early AI work in transportation leaned on operations research and rule-based expert systems for tasks such as vehicle routing, timetable construction, and airline crew scheduling. The shift toward statistical and learning-based methods accelerated in the 2010s as deep learning matured. Convolutional neural networks made reliable visual perception possible, recurrent and later transformer architectures improved time-series forecasting, and reinforcement learning offered a framework for sequential control problems such as signal timing and fleet repositioning. The DARPA Grand and Urban Challenges of 2004 to 2007 catalyzed the autonomous-driving industry specifically; that lineage is covered in the autonomous driving article.
AI appears across every major mode of transport, with different maturity levels in each. The table below summarizes the principal application areas discussed in this article.
| Mode | Representative AI applications | Maturity as of 2026 |
|---|---|---|
| Road (vehicles) | Autonomous driving, ADAS (lane keeping, automatic braking), driver monitoring | Level 2 widespread; Level 4 robotaxis commercial in selected cities |
| Road (network) | Adaptive signal timing, congestion and incident prediction, tolling | Widely deployed; learning-based optimization expanding |
| Logistics and freight | Route optimization, freight matching, demand and inventory forecasting, warehouse robotics | Mature and economically significant |
| Ride-hailing | Demand prediction, dynamic pricing, dispatch and matching, ETA prediction | Mature, in daily production use |
| Public transit | Ridership forecasting, schedule and crew optimization, on-demand microtransit | Growing |
| Rail | Predictive maintenance, automatic train operation, defect detection | Operational at varying grades of automation |
| Aviation | Flight-path and fuel optimization, predictive maintenance, weather-aware traffic flow | Decision-support tools in production; full automation limited |
| Maritime | Route and fuel optimization, collision-avoidance research, autonomous vessels | Mostly pilots and constrained routes |
| Last-mile delivery | Sidewalk and road delivery robots, drone delivery | Commercial at growing scale |
The most visible application of AI to road vehicles is automated driving. Systems are commonly classified using the SAE J3016 taxonomy of six levels, from Level 0 (no automation) to Level 5 (full automation under all conditions). Most vehicles sold with automation today operate at Level 2, where the system controls steering and speed but a human must supervise continuously; examples include Tesla Autopilot, GM Super Cruise, and Ford BlueCruise. Level 4 systems, which drive without human intervention inside a defined operational area, are deployed commercially as robotaxis by operators such as Waymo and Baidu's Apollo Go, and in autonomous trucking by Aurora. No commercially deployed system had achieved Level 5 as of 2026.3
Below full autonomy, advanced driver-assistance systems (ADAS) apply the same underlying AI techniques, computer vision and sensor fusion, to features such as automatic emergency braking, adaptive cruise control, lane-keeping assistance, blind-spot warning, and driver-attention monitoring. These features are now common on mainstream vehicles and represent the largest real-world footprint of transportation AI on the road, even though they fall short of self-driving.
Because the technical stack, sensors, key companies, regulation, and safety record of automated vehicles are covered in depth elsewhere, this article does not duplicate that material. See the autonomous driving article for perception pipelines, LiDAR versus vision debates, robotaxi deployments, the SAE levels in detail, and the history of accidents and investigations.
Urban road networks are a natural target for AI because small improvements in signal timing aggregate into large reductions in delay, fuel use, and emissions. Beyond the classical adaptive systems (SCATS, SCOOT), recent work applies machine learning and reinforcement learning to traffic-signal control, treating intersections as agents that learn timing policies from observed flow. Research deployments and simulations have reported meaningful reductions in vehicle delay relative to fixed-time plans, though results vary with traffic conditions and some studies find limited benefit where traffic is already predictable.4
A prominent industrial example is Google's Project Green Light. Launched in 2023, it uses aggregated Google Maps driving trends to model how traffic flows through an intersection and recommends adjustments to signal timing that city engineers can apply using existing equipment, without new hardware. Google states the program was live in 20 cities across four continents, citing example deployments including Haifa, Kolkata, Hamburg, and Boston, and reports the potential for up to a 30% reduction in stops and a 10% reduction in greenhouse gas emissions at coordinated intersections, with the company noting these figures are averaged across coordinated intersections and expected to evolve.5 In Boston, the city said it applied Green Light recommendations across more than one hundred intersections.6 As with any model-based recommendation system, the benefits depend on local conditions and on cities correctly implementing the changes.
AI is also used for traffic and incident prediction, estimating where congestion or crashes are likely so that operators can respond proactively, and for camera-based detection of incidents, wrong-way drivers, and queue lengths. Consumer navigation apps such as Google Maps and Waze use machine learning over fleet and crowd-sourced data to predict travel times and route around congestion.
Logistics is among the most economically significant and mature areas of transportation AI. The core problems, deciding which vehicle carries which load and in what sequence to make stops, are variants of the vehicle routing problem, which is computationally hard and well suited to optimization combined with machine-learning-based demand forecasting.
The best-known operational example is UPS's ORION (On-Road Integrated Optimization and Navigation), first deployed at scale in the 2010s. UPS states that ORION optimizes routes by considering package details, delivery commitments, and road conditions, and that the system saves the company roughly 100 million miles and about 10 million gallons of fuel per year, on the order of 300 to 400 million dollars in annual cost, and roughly 100,000 metric tons of carbon-dioxide emissions, across tens of thousands of routes.7 These are company-reported figures. The wider market for AI in logistics has grown quickly, with freight brokers and carriers using machine learning for freight matching (pairing loads with available trucks), dynamic pricing, warehouse robotics, and demand and inventory forecasting.8
Forecasting is central: retailers and carriers use time-series and machine-learning models to anticipate shipment volumes, position inventory, and plan capacity, particularly around seasonal peaks. Within warehouses, robotics and computer vision handle picking, sorting, and movement, complementing the routing intelligence that governs the vehicles outside.
Ride-hailing platforms such as Uber and Lyft run several production machine-learning systems that together coordinate supply and demand in real time. The principal applications are:
These systems illustrate how AI in transportation often operates as coordination and forecasting infrastructure rather than as a physically autonomous machine.
Public transit agencies apply AI to ridership forecasting, schedule and crew optimization, and real-time arrival prediction. Demand models help match service frequency to need and plan network changes. On-demand microtransit, where small vehicles are dispatched dynamically in response to requests rather than running fixed routes, relies on routing and matching algorithms similar to those used in ride-hailing. Computer vision is also used for passenger counting and for monitoring station and platform safety. Many transit applications remain at the pilot or partial-deployment stage, constrained by data availability, procurement cycles, and the operational caution appropriate to safety-critical public services.
Railways apply AI in three main ways: predictive maintenance, automatic train operation, and inspection.
Predictive maintenance uses sensor data from locomotives, rolling stock, and track to anticipate component failures before they occur, reducing service disruption. Surveys of the field describe a move toward digital twins and edge AI for monitoring railway infrastructure.11 Computer vision is used to inspect track, overhead lines, and rolling stock for defects, automating tasks that were traditionally manual.
Automatic train operation (ATO) is described using grades of automation from GoA1 to GoA4, where GoA4 denotes unattended operation with no staff on board. Many metro lines run at high grades of automation in closed, signal-protected environments. The most cited heavy-freight example is Rio Tinto's AutoHaul in the Pilbara region of Western Australia, which the company describes as the world's first fully autonomous, long-distance, heavy-haul rail network. AutoHaul became fully operational in June 2019; Rio Tinto has reported operating roughly 200 locomotives across more than 1,700 kilometres of track, carrying iron ore from 16 mines to four port terminals, under a programme it valued at about 940 million US dollars.12 Autonomy is more tractable on rail than on open roads because trains run on fixed guideways with centralized signaling, which constrains the perception and decision problem.
In aviation, AI is used today mainly as decision support rather than as a replacement for pilots or air-traffic controllers, both of which remain human roles under strict regulation.
Flight-path and fuel optimization is a established commercial application. Alaska Airlines was an early adopter of the Flyways platform from Airspace Intelligence, which uses machine learning to analyze weather, winds, turbulence, airspace constraints, and traffic volume and to recommend more efficient routes to dispatchers. Alaska reported that optimized routes saved over 1.2 million gallons of fuel in a year, equivalent to about 11,958 metric tons of carbon dioxide, with average fuel and emissions savings of three to five percent on flights longer than four hours; the airline emphasized the tool supports its dispatchers rather than replacing them.13
Predictive maintenance for aircraft engines and systems, drawing on sensor telemetry, is widely used by airlines and manufacturers to reduce unscheduled downtime. In air-traffic management, agencies are integrating better weather modeling and traffic-flow tools; the United States Federal Aviation Administration's NextGen modernization program includes an automated NextGen Weather Processor that supports strategic traffic-flow management by translating weather information into predicted airspace constraints.14 Direct AI control of separation between aircraft remains largely at the research stage, given the safety-critical and heavily regulated nature of the task. Autonomous drones (uncrewed aircraft) are a faster-moving area, with AI handling perception, navigation, and obstacle avoidance for inspection, surveying, and parcel delivery; see last-mile delivery below.
In the maritime sector, AI supports voyage and fuel optimization, engine and hull condition monitoring, port operations, and a growing body of research on autonomous collision avoidance. Weather routing and trim optimization use models to reduce fuel burn over long voyages.
Fully autonomous vessels remain limited to pilots and constrained routes. A frequently cited example is the MV Yara Birkeland, an electric container ship in Norway designed to carry fertilizer over a short coastal route, with development focused on situational awareness and collision-avoidance systems built to high safety-integrity standards. Other examples include the autonomous ferry Falco and the container ship Zhi Fei.15 A central challenge for autonomous ships is performing collision avoidance in a way that complies with the international COLREG navigation rules, which were written for human watchkeepers; much current AI research targets exactly this problem.16
The final leg of delivery has become a distinct AI application area, using small autonomous robots on sidewalks and roads as well as aerial drones. These systems run perception and navigation stacks conceptually similar to those in self-driving cars, but at lower speeds and smaller scale, which lowers risk.
Starship Technologies operates one of the largest sidewalk-robot fleets. The company announced that its robots surpassed 8 million autonomous deliveries in April 2025, operating a fleet it described as more than 2,000 robots across more than 150 locations in six countries at Level 4 autonomy, having traveled over 10 million delivery miles, with a strong presence on university campuses.17 Serve Robotics, which was spun out of Uber in 2021, reported deploying its 1,000th third-generation robot in October 2025 and targeting 2,000 robots by year-end, delivering for partners including Uber Eats, DoorDash, and 7-Eleven; in 2025 it acquired Vayu Robotics to strengthen its AI navigation models.18 Nuro, which built road-going autonomous delivery vehicles, announced in 2024 a pivot toward licensing its autonomous-driving system to other companies, and in 2025 entered a program with Lucid and Uber to bring the Nuro Driver to a robotaxi fleet.19
A common set of AI methods underlies these diverse applications.
| Technique | Role in transportation |
|---|---|
| Computer vision | Detecting and classifying vehicles, pedestrians, lane markings, track and infrastructure defects, and reading signs and signals |
| Sensor fusion | Combining cameras, LiDAR, radar, GPS, and inertial data into a single environment model |
| Deep learning | The basis of modern perception and many forecasting and control models |
| Forecasting and time-series models | Predicting demand, travel times, congestion, and maintenance needs |
| Reinforcement learning | Sequential decision making such as signal control, fleet repositioning, and motion planning |
| Optimization and operations research | Vehicle routing, scheduling, crew assignment, and network flow |
Computer vision provides the perception layer for vehicles, robots, and inspection systems, using convolutional neural networks and transformer-based detectors for object detection, segmentation, and tracking. Sensor fusion integrates complementary sensors so that the weaknesses of one (for example, a camera's difficulty estimating depth, or LiDAR's degradation in heavy rain) are offset by others. Forecasting models, from gradient-boosted trees to recurrent and transformer networks, predict demand, travel times, and failures. Reinforcement learning frames control problems as agents learning policies through interaction, applied to signal timing, vehicle dispatch, and motion planning, though deployment in safety-critical settings is cautious. Underpinning much of the recent progress is the trend toward larger foundation models trained on large driving and operational datasets, which can improve generalization across environments.
The table below lists representative systems across modes. Performance figures attributed to a company are that company's own claims unless noted; independent verification varies by case.
| System or company | Mode | Application | Notes |
|---|---|---|---|
| Waymo (Alphabet) | Road | Level 4 robotaxi | Commercial service in several US cities; see autonomous driving |
| Tesla | Road | Level 2 driver assistance and robotaxi pilot | Vision-based approach; see autonomous driving |
| Aurora | Road (freight) | Autonomous trucking | Driverless freight runs in Texas |
| Project Green Light (Google) | Road network | Signal-timing recommendations | 20 cities, four continents (Google)5 |
| UPS ORION | Logistics | Delivery route optimization | ~100M miles saved per year (UPS)7 |
| Uber / Lyft | Ride-hailing | Demand prediction, pricing, dispatch, ETA | DeepETA and ML pricing10 |
| Rio Tinto AutoHaul | Rail (freight) | GoA4 autonomous heavy-haul | Fully operational 2019, Pilbara (Rio Tinto)12 |
| Flyways (Airspace Intelligence) | Aviation | Flight-path optimization | Used by Alaska Airlines13 |
| MV Yara Birkeland | Maritime | Autonomous electric container ship | Short coastal route in Norway15 |
| Starship Technologies | Last-mile | Sidewalk delivery robots | 8M+ deliveries by April 2025 (Starship)17 |
| Serve Robotics | Last-mile | Sidewalk delivery robots | Spun out of Uber; partners include DoorDash18 |
Proponents argue that AI in transportation can deliver three broad categories of benefit.
Safety. Road crashes are a major cause of death worldwide; the World Health Organization estimated about 1.19 million road-traffic deaths per year in its 2023 global status report.20 Driver-assistance features such as automatic emergency braking are intended to reduce crashes, and operators of automated vehicles publish safety data in support of the technology. Waymo, for example, has published peer-reviewed analyses and an insurance-data study (conducted with Swiss Re) reporting lower crash and liability-claim rates for its vehicles than for human-driven benchmarks over tens of millions of miles.21 These are encouraging but contested claims; see the discussion of risks below.
Efficiency. Routing, signal timing, and flight-path optimization reduce wasted distance, idling, and delay. The UPS ORION and Alaska Flyways figures above illustrate the scale of potential savings in fuel and time, and adaptive signal control can cut intersection delay relative to fixed schedules.
Emissions and energy. Because fuel use is closely tied to distance, idling, and routing, efficiency gains translate into lower emissions. Project Green Light, ORION, and Flyways each frame their benefits partly in terms of avoided carbon-dioxide emissions, and demand prediction can improve vehicle utilization so that fewer empty miles are driven.
The deployment of AI in transportation raises significant concerns.
Safety incidents and the limits of safety claims. Automated systems have been involved in serious crashes. The 2018 death of a pedestrian struck by an Uber test vehicle in Tempe, Arizona, and a 2023 incident in which a Cruise robotaxi dragged a pedestrian in San Francisco, both prompted investigations and regulatory action; both are detailed in the autonomous driving article. Critics caution that company safety statistics often reflect operation in favorable conditions and well-mapped areas, and that comparing automated-vehicle crash rates to human averages can mislead without controlling for those conditions. Company-reported figures should be read as claims rather than independent findings.
Liability and accountability. When an automated system causes harm, responsibility may shift from a human driver to a manufacturer, software provider, or fleet operator, raising unsettled questions for law and insurance. The Cruise case, in which the company was found to have submitted a false report to a federal regulator, underscored the importance of transparent incident reporting.
Jobs. Automation threatens driving and operating jobs. Truck driving is one of the most common occupations in the United States, with roughly 3.5 million drivers, a figure cited by the American Trucking Associations.22 Autonomous trucking, delivery robots, and automated rail and ports could displace some of these roles over time even as they create new ones in technology and maintenance; the pace and net effect are debated.
Equity and access. Benefits and harms may be distributed unevenly. Robotaxi and delivery services often launch in dense, well-mapped, affluent areas first. There are also concerns about how perception systems perform across different pedestrian appearances and conditions, and about the effect of new services on existing public transit and on curb space.
Security and data. Connected and automated transportation systems introduce cybersecurity risks, and the data used for demand prediction and routing raises privacy questions. International vehicle regulations now include cybersecurity and software-update requirements for connected and automated vehicles.
Regulation of transportation AI is fragmented across modes and jurisdictions and is generally mode-specific.
For road vehicles, the United States relies on a mix of federal vehicle-safety oversight by the National Highway Traffic Safety Administration (NHTSA) and a patchwork of state laws governing testing and deployment; internationally, the United Nations Economic Commission for Europe coordinates harmonized vehicle regulations, including rules on automated lane-keeping and on cybersecurity and software updates. The regulatory landscape for automated driving specifically is covered in the autonomous driving article.
Other modes are governed by their established safety regulators: aviation by bodies such as the FAA and the European Union Aviation Safety Agency, which treat AI primarily as decision-support subject to existing certification regimes; rail by national rail-safety authorities; and shipping by the International Maritime Organization, whose COLREG rules constrain how autonomous vessels may navigate. A cross-cutting issue is that most of these frameworks were written for human operators and are being adapted incrementally to accommodate automated systems.
As of 2026, AI is firmly established in transportation as forecasting, optimization, and decision-support infrastructure, and is expanding in physical autonomy at a measured pace. The most mature applications, logistics routing, ride-hailing coordination, and adaptive traffic management, operate at large scale and deliver measurable efficiency gains. Physical autonomy is commercial but geographically bounded: Level 4 robotaxis and autonomous trucks run in selected regions, sidewalk delivery robots operate in millions of deliveries per year, and autonomous heavy-haul rail is in routine service in constrained settings, while fully autonomous shipping and pilotless commercial aviation remain largely at the research or pilot stage.
Historically, timelines for full autonomy have been optimistic, and industry expectations have repeatedly slipped. The more reliable near-term trajectory is incremental: wider driver assistance, broader robotaxi and delivery footprints, more pervasive optimization behind the scenes, and continued integration of larger learned models, accompanied by regulation that adapts mode by mode. Whether AI ultimately makes transportation substantially safer and cleaner will depend not only on the technology but on how transparently its performance is measured and how its benefits and disruptions are distributed.
Federal Highway Administration and related ITS literature on adaptive signal control history (SCATS, SCOOT); SCATS deployment in Troy, Michigan (June 1992). U.S. DOT / FHWA, "Adaptive Signal Control." https://www.fhwa.dot.gov/publications/research/randt/evaluations/17007/17007.pdf Accessed 2026-05-31. ↩
U.S. DOT ITS Joint Program Office, "History of Intelligent Transportation Systems." https://www.itsga.org/wp-content/uploads/2016/08/ITS-JPO-History-of-ITS.pdf Accessed 2026-05-31. ↩
SAE International, "SAE J3016: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles" (2021). See also the autonomous driving article. Accessed 2026-05-31. ↩
Survey and evaluation literature on adaptive and reinforcement-learning traffic signal control, e.g. Federal Highway Administration evaluations and academic reviews of adaptive signal control benefits and limitations. https://www.fhwa.dot.gov/publications/research/randt/evaluations/17007/17007.pdf Accessed 2026-05-31. ↩
Google Research, "Green Light." https://sites.research.google/gr/greenlight/ Accessed 2026-05-31. Figures (20 cities; up to 30% fewer stops and 10% lower emissions at coordinated intersections) are Google's stated claims, qualified by the company as averages expected to evolve. ↩ ↩2
City of Boston, "Mayor Wu Announces Expansion of Project Green Light Signal Optimization Program." https://www.boston.gov/news/mayor-wu-announces-expansion-project-green-light-signal-optimization-program Accessed 2026-05-31. ↩
UPS describes ORION (On-Road Integrated Optimization and Navigation) and its reported savings (approximately 100 million miles and 10 million gallons of fuel per year; roughly 300 to 400 million dollars; about 100,000 metric tons of CO2). Figures are UPS company claims as reported in UPS materials and industry case studies. https://www.bsr.org/en/case-studies/center-for-technology-and-sustainability-orion-technology-ups Accessed 2026-05-31. ↩ ↩2
Industry reporting on AI adoption in logistics and supply chain, including freight matching and forecasting, 2025. Accessed 2026-05-31. ↩
Reporting and technical descriptions of ride-hailing demand prediction and dynamic pricing at Uber and Lyft. Accessed 2026-05-31. ↩ ↩2
Uber Engineering, "DeepETA: How Uber Predicts Arrival Times Using Deep Learning," describing a transformer-based ETA model. https://www.uber.com/blog/deepeta-how-uber-predicts-arrival-times/ Accessed 2026-05-31. ↩ ↩2
"A Survey of AI-Enabled Predictive Maintenance for Railway Infrastructure: Models, Data Sources, and Research Challenges." https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12899784/ Accessed 2026-05-31. ↩
Rio Tinto, "World-first autonomous trains deployed at Rio Tinto's iron ore operations" (2018) and related releases; "Successful rollout of AutoHaul." Fully operational June 2019; ~200 locomotives over 1,700+ km, 16 mines, four ports; ~US$940 million programme. https://www.riotinto.com/en/news/releases/2018/world-first-autonomous-trains-deployed Accessed 2026-05-31. ↩ ↩2
Alaska Airlines, "How AI is helping Alaska Airlines plan better flight routes and lower emissions." Over 1.2 million gallons of fuel saved in a year (~11,958 metric tons CO2); 3 to 5 percent savings on flights longer than four hours. https://news.alaskaair.com/sustainability/how-ai-is-helping-alaska-airlines-plan-better-flight-routes-and-lower-emissions/ Accessed 2026-05-31. ↩ ↩2
Federal Aviation Administration, NextGen and NextGen Weather Processor. https://www.faa.gov/nextgen/programs/weather Accessed 2026-05-31. ↩
"MV Yara Birkeland," autonomous electric container ship in Norway; coverage of autonomous-vessel development and examples (Falco, Zhi Fei). https://en.wikipedia.org/wiki/MV_Yara_Birkeland Accessed 2026-05-31. ↩ ↩2
Research literature on autonomous-ship collision avoidance and compliance with the COLREG navigation rules. Accessed 2026-05-31. ↩
Starship Technologies, "Starship Technologies Surpasses 8 Million Deliveries" (April 2025); fleet of 2,000+ robots across 150+ locations in six countries at Level 4 autonomy. https://www.starship.xyz/press/starship-technologies-surpasses-8-million-deliveries/ Accessed 2026-05-31. ↩ ↩2
Serve Robotics, SEC Form 8-K filings (2025) reporting deployment of its 1,000th third-generation robot (October 2025), a target of 2,000 robots by year-end, partnerships including DoorDash, and the acquisition of Vayu Robotics. https://www.sec.gov/Archives/edgar/data/0001832483/000183248325000107/serv-20250930xex991earning.htm Accessed 2026-05-31. ↩ ↩2
Reporting on Nuro's 2024 pivot to licensing its autonomous-driving system and its 2025 program with Lucid and Uber. https://en.wikipedia.org/wiki/Nuro Accessed 2026-05-31. ↩
World Health Organization, "Global status report on road safety 2023" (approximately 1.19 million road-traffic deaths per year). https://www.who.int/news/item/13-12-2023-despite-notable-progress-road-safety-remains-urgent-global-issue Accessed 2026-05-31. ↩
Waymo, "New Swiss Re study: Waymo is safer than even the most advanced human-driven vehicles" (2024) and peer-reviewed crash-rate analysis ("Comparison of Waymo Rider-Only Crash Rates by Crash Type to Human Benchmarks," Traffic Injury Prevention, 2025). Figures are from Waymo and associated studies. https://waymo.com/blog/2024/12/new-swiss-re-study-waymo/ Accessed 2026-05-31. ↩
American Trucking Associations workforce data, reporting roughly 3.5 million truck drivers in the United States. https://www.trucking.org/news-insights/ata-releases-updated-driver-shortage-report-and-forecast Accessed 2026-05-31. ↩