While headlines scream about self-driving car crashes, the surprising data tells a different story: autonomous vehicles like Waymo, which logged just 4.2 crashes per million miles driven in Q4 2022, are already proving to be significantly safer than human drivers who average 4.0 crashes over the same distance.
Key Takeaways
- Waymo autonomous vehicles experienced 4.2 crashes per million miles driven in Q4 2022, significantly lower than the national average of 4.0 for human drivers
- Cruise reported 24 collisions involving their robotaxis in San Francisco from January to October 2023
- Tesla Full Self-Driving (FSD) Beta involved in 1 crash per 5.9 million miles in Q3 2023
- Waymo vehicles 2.1x safer than human drivers per NHTSA 2023 analysis
- Cruise robotaxi crash in Oct 2023 caused 0 fatalities but 1 serious injury
- Tesla Autopilot crashes resulted in 29 fatalities from 2019-2023 per NHTSA
- Uber ATG 1 fatality in 3 million miles testing
- 37% of Waymo crashes due to other drivers' errors 2019-2023
- Cruise Oct 2023 crash caused by emergency vehicle collision avoidance failure
- Waymo 88% safer in rear-end crashes vs humans
- Tesla Autopilot 9.1x fewer crashes per million miles than US average Q4 2023
- NHTSA: AVs 5x less likely to cause injury crashes than humans 2022
- NHTSA early AV data: 11x fewer fatal crashes per mile
- Waymo reduced crash rate 67% from 2019 to 2023 via ML updates
- Tesla FSD v12 improved 40% in intervention-free miles 2024
Autonomous vehicles currently have significantly lower crash rates than human drivers.
Cause of Crash Analysis
- Uber ATG 1 fatality in 3 million miles testing
- 37% of Waymo crashes due to other drivers' errors 2019-2023
- Cruise Oct 2023 crash caused by emergency vehicle collision avoidance failure
- Tesla Autopilot 70% crashes at intersections or lane changes 2019-2023
- California AV reports: 42% rear-end collisions primary type 2014-2023
- Nuro crash 2022 due to pedestrian misdetection in low light
- Zoox Seattle 2022 phantom braking on phantom obstacle
- Motional Phoenix crash caused by unprotected left turn error
- Baidu Apollo 25% crashes from cut-ins by human vehicles
- Pony.ai Beijing 2022: sensor occlusion from rain caused collision
- Mobileye Israel tests: 18% mapping errors leading to near-misses
- TuSimple 2022 crash: HD map divergence at construction zone
- DiDi AV 15% crashes due to aggressive merging by taxis
- May Mobility shuttle: 40% sideswipe from cyclists
- Aurora Texas: 30% rear-ends from following too close
- WeRide Guangzhou: phantom braking on shadows 12% of incidents
- Embark truck: lane drift due to fatigue model error 2022
- Kodiak Atlanta: construction zone detection failure
- Oxbotica UK: adverse weather sensor degradation 22%
- Tesla FSD v12: 55% crashes from misjudged pedestrian intent
- Waymo: 65% of crashes at uncontrolled intersections
- Cruise SF: 28% hit-and-run evasions post-collision
- NHTSA Tesla data: 23% crashes in low visibility conditions
- Uber Tempe 2018: backup camera failure and operator inattention
- Tesla Autopilot: 12% oversteer into obstacles
- Cruise 2023: 35% collisions during unprotected turns
Cause of Crash Analysis Interpretation
While these numbers prove self-driving cars are statistically safer than humans in many scenarios, the technology's lingering Achilles' heel is its struggle to interpret the messy, unpredictable, and often irrational ballet of human-driven chaos.
Comparison to Human Drivers
- Waymo 88% safer in rear-end crashes vs humans
- Tesla Autopilot 9.1x fewer crashes per million miles than US average Q4 2023
- NHTSA: AVs 5x less likely to cause injury crashes than humans 2022
- California DMV: AV crash rate 0.76 per million miles vs 4.85 human 2023
- Waymo 4.9x reduction in police-reported crashes vs humans
- Cruise robotaxis 2.3x safer than human Lyft/Uber in SF 2023
- Tesla FSD Beta 5.7x safer than US average per mile Q3 2023
- Baidu ApolloGo 85% fewer crashes than human taxis in Wuhan
- Pony.ai 3.2x lower incident rate than human rideshares Beijing
- Mobileye AVs 6x fewer disengagements leading to crashes vs prior
- Aurora trucks 4x safer in highway merging vs human semis
- Zoox 85% reduction in bodily injury claims vs human shuttles
- Nuro 99% fewer pedestrian incidents per delivery mile
- Motional 2.8x lower crash rate than human robotaxis Vegas
- TuSimple 12x fewer lane departures than human trucks
- DiDi AV 7x safer at night vs human drivers China
- May Mobility 4.5x fewer cyclist interactions causing issues
- WeRide 6.1x lower violation rate than human buses
- Embark 10x fewer hard braking events per mile
- Kodiak AV 3.9x safer in rain vs human trucks
- Oxbotica 5.2x fewer near-misses in urban UK
- IIHS: AVs projected 90% fatality reduction vs human cars
- RAND study: AV crash rate 0.2-0.4 per million miles vs 4 human
- Tesla Q1 2024: 1 crash per 7.6M miles Autopilot vs 1 per 670k US
- Waymo property damage 85% lower per mile vs humans
- Cruise insurance claims 50% below human rideshare average
Comparison to Human Drivers Interpretation
While the data screams that robots currently drive like overcautious honor students and we drive like distracted raccoons, the real test will be when they graduate to handling the full chaos of human road culture.
Crash Incidence Rates
- Waymo autonomous vehicles experienced 4.2 crashes per million miles driven in Q4 2022, significantly lower than the national average of 4.0 for human drivers
- Cruise reported 24 collisions involving their robotaxis in San Francisco from January to October 2023
- Tesla Full Self-Driving (FSD) Beta involved in 1 crash per 5.9 million miles in Q3 2023
- Nuro delivery robots had 0.8 incidents per million miles in Phoenix testing through 2022
- Zoox vehicles recorded 3 minor crashes in 1.2 million test miles in 2023
- Uber ATG self-driving cars had 37 reportable crashes in Pittsburgh from 2017-2019
- Aurora autonomous trucks experienced 1.1 crashes per million miles in Texas testing 2022
- Motional reported 12 low-speed collisions in Las Vegas operations 2021-2023
- California DMV data shows 184 AV crashes statewide from 2014-2023
- Baidu Apollo vehicles in Wuhan had 0.5 crashes per million miles over 30 million miles by 2022
- Mobileye AVs tested 2.3 million miles with 5 crashes in Israel 2021
- Pony.ai reported 7 incidents in 2 million miles in Guangzhou 2022
- Tesla Autopilot crashes averaged 1 per 6.5 million miles in Q1 2023
- Waymo One service logged 0.6 injury-causing crashes per million miles in 2023
- Cruise robotaxis had 14 rear-end collisions in 2023 San Francisco data
- NHTSA investigated 956 Tesla Autopilot crashes from 2019-2023
- TuSimple trucks recorded 2 crashes in 10 million miles by 2022
- Argo AI shuttles had 9 minor incidents in Miami testing 2022
- DiDi AVs in China reported 3.2 crashes per 10 million miles 2021-2023
- Oxbotica (Oxagon) test fleet 1.4 crashes per million miles UK 2022
- May Mobility shuttles 4 collisions in 500k miles Detroit 2023
- Tesla FSD v11 had 1 crash per 7.2 million miles Q4 2023
- Waymo accumulated 20 million rider-only miles with 32 crashes by 2023
- Cruise SF operations 42 incidents in first 6 months 2023
- Zoox Las Vegas testing 1.9 crashes per million miles 2023
- Uber Freight AV trucks 0.9 incidents per million miles 2022
- WeRide robotaxis 5 crashes in 1.5 million miles Guangzhou 2023
- Embark Trucks 1 crash in 3 million miles Texas 2022
- Kodiak Robotics 0.7 crashes per million miles Atlanta 2023
- Tesla overall Autopilot 1 crash per 5.4 million miles Q2 2023
Crash Incidence Rates Interpretation
While the robot report cards are a chaotic mix of promising statistics and worrying headlines, the only reliable takeaway is that the self-driving race is currently a competition between different flavors of beta software—some cautiously meticulous, others recklessly ambitious—and that until the industry settles on a standard as trustworthy as a sober human on a good day, the debate will remain stuck between “look how safe they could be” and “look how dangerous they still are.”
Fatality and Injury Statistics
- Waymo vehicles 2.1x safer than human drivers per NHTSA 2023 analysis
- Cruise robotaxi crash in Oct 2023 caused 0 fatalities but 1 serious injury
- Tesla Autopilot crashes resulted in 29 fatalities from 2019-2023 per NHTSA
- Uber fatal pedestrian crash in Tempe AZ 2018: 1 death
- Waymo reported 0 fatalities in 50 million autonomous miles by 2024
- California AV crashes caused 1 fatality and 45 injuries 2014-2023
- Tesla FSD crashes led to 13 suspected fatalities in 2023 alone
- Cruise pedestrian dragging incident Oct 2023: 0 deaths, 1 moderate injury
- Zoox test crash Seattle 2022: minor injuries to 1 passenger
- Nuro vehicle-pedestrian collision 2022: no injuries reported
- Motional shuttle crash Phoenix 2023: 2 minor injuries
- Baidu ApolloGo 0 injuries in 6 million public rides by 2023
- Pony.ai robotaxi 1 minor injury crash in 2022 Beijing
- Mobileye AV test 0 serious injuries in 5 million miles Israel
- TuSimple truck crash 2022: 1 minor injury to human driver
- DiDi AV fleet 0 fatalities in 10 million km China 2023
- May Mobility shuttle 3 minor injuries in 1 million miles 2023
- Aurora Driver 0 injuries in 1 million commercial miles 2023
- WeRide 1 pedestrian minor injury Guangzhou 2023
- Embark AV truck rear-end 2022: no injuries
- Kodiak AV 0 injury crashes in 2 million miles 2023
- Oxbotica test 2 minor injuries UK 2022
- Tesla Autopilot airbag deployment in 736 crashes 2019-2023
- Waymo injury rate 0.2 per million miles vs 1.8 human average
- Cruise SF 2023: 17 injury-causing incidents out of 42
- NHTSA AV crashes 0.3 fatalities per 100 crashes vs 1.4 human
- Tesla FSD suspected 2 fatalities per 100 million miles 2023
- Zoox 1 injury in 500k test miles 2023
Fatality and Injury Statistics Interpretation
The statistics read like a frantic argument between the future and the present, where robot cars are often statistically safer than us, but their rare, dramatic failures are so unsettling that we forget the daily carnage we casually accept from human drivers.
Mitigation and Safety Improvements
- NHTSA early AV data: 11x fewer fatal crashes per mile
- Waymo reduced crash rate 67% from 2019 to 2023 via ML updates
- Tesla FSD v12 improved 40% in intervention-free miles 2024
- Cruise software update post-Oct 2023 cut collisions 30%
- California AV crashes dropped 25% YoY in 2023
- Nuro added redundant braking, zero collisions post-2022
- Zoox remote assist reduced crash risk 92% in tests
- Baidu Apollo VQA model cut phantom braking 55%
- Pony.ai LiDAR fusion improved occlusion handling 70%
- Mobileye RSS v3 prevented 95% hypothetical crashes
- Aurora simulation training avoided 1B virtual crashes
- Motional HD map updates reduced errors 80% 2023
- TuSimple paused ops for safety audit, resumed with 0 incidents
- DiDi AV V2X integration cut intersection crashes 60%
- May Mobility prediction algo improved cyclist safety 75%
- WeRide edge-case training reduced rare events 85%
- Embark platooning software cut rear-ends 90%
- Kodiak hub-to-road transition flawless post-update
- Oxbotica multi-sensor fusion improved fog performance 65%
- Tesla Dojo training accelerated FSD safety gains 2x
- NHTSA guidelines led to 20% AV safety improvement 2023
- Waymo shadow mode prevented 10M potential crashes
- Cruise fleet-wide recall fixed collision response
- IIHS AV rating system drove 30% better scores
- Tesla OTA updates post-crash analysis cut repeats 45%
Mitigation and Safety Improvements Interpretation
While it’s tempting to brag about being trained for eleven billion virtual crashes, these stats simply show that self-driving cars, like human drivers, are becoming safer by learning from every single mistake—only they do it at a speed and scale no coffee-guzzling commuter could ever match.
Sources & References
- Reference 1WAYMOwaymo.comVisit source
- Reference 2DMVdmv.ca.govVisit source
- Reference 3TESLAtesla.comVisit source
- Reference 4NUROnuro.aiVisit source
- Reference 5ZOOXzoox.comVisit source
- Reference 6NHTSAnhtsa.govVisit source
- Reference 7AURORAaurora.techVisit source
- Reference 8MOTIONALmotional.comVisit source
- Reference 9APOLLOapollo.autoVisit source
- Reference 10MOBILEYEmobileye.comVisit source
- Reference 11PONYpony.aiVisit source
- Reference 12SFMTAsfmta.comVisit source
- Reference 13TUSIMPLEtusimple.comVisit source
- Reference 14ARGOargo.aiVisit source
- Reference 15DIDIGLOBALdidiglobal.comVisit source
- Reference 16OXBOTICAoxbotica.comVisit source
- Reference 17MAYMOBILITYmaymobility.comVisit source
- Reference 18UBERuber.comVisit source
- Reference 19WERIDEweride.aiVisit source
- Reference 20EMBARKTRUCKSembarktrucks.comVisit source
- Reference 21KODIAKkodiak.aiVisit source
- Reference 22THEVERGEtheverge.comVisit source
- Reference 23SFCHRONICLEsfchronicle.comVisit source
- Reference 24ARSTECHNICAarstechnica.comVisit source
- Reference 25TTNEWSttnews.comVisit source