Uber

A test vehicle, not a production car

The vehicle in the Tempe crash of 18 March 2018 is described as an Uber test vehicle and a prototype Uber self-driving car^[1][2]. It was a 2017 Volvo XC90 that Uber had modified: the car carried a single lidar unit on the forward half of its roof and radar sensors giving a 360-degree view of its surroundings^[1][2]. Those modified XC90s did not include any of Volvo's own advanced driver-assistance systems^[1][2].

The distinction matters for understanding what the car could and could not do. Lidar measures distance by bouncing laser light off objects and builds a point-by-point picture of the road ahead; radar does a similar job with radio waves, which works well at long range and in poor weather. Advanced driver-assistance systems are the features familiar from production cars — warning of an imminent collision, braking on their own, keeping a car inside its lane. Uber's prototypes were stripped of that manufacturer layer of protection and relied on Uber's own software instead^[1][2].

Computer control and a disabled function

While the vehicle was under computer control, emergency braking manoeuvres were not enabled, a choice intended to reduce the potential for erratic vehicle behaviour^[1][2]. In other words, the automated system was built to steer and to warn but not to perform the hardest braking itself; that action was left to the person in the driver's seat.

That person is the safety driver, a standard feature of road testing of automated systems: a trained human who sits behind the wheel, watches the system's output, and is expected to take over when the software reaches the limit of what it can handle. The arrangement assumes the human is attentive at the moment the machine fails.

The crash and the investigation

The collision was fatal^[1]. The National Transportation Safety Board (NTSB) sent a team of federal investigators to gather data from the vehicle's instruments and to examine the condition of the vehicle and the actions of the safety driver^[1].

In an interview after the crash, the safety driver told the NTSB that she had been monitoring system messages on the centre console and had not used either of her cell phones until she called 911^[1]. On 21 March 2018, Tempe police released video showing the safety driver looking down in the period before the collision^[1].

Findings and recommendations

The NTSB noted that the nascent computerised braking technology was disabled on the day of the crash, and that human intervention was still legally required^[1][2]. It issued a series of recommendations and sharply criticised Uber^[1][2].

Why it mattered

The Arizona incident magnified the importance of collision avoidance systems for self-driving vehicles^[1]. Collision avoidance is the last line of defence in any driving system, human or machine: the layer that acts when everything upstream — perception, prediction, planning — has already failed. The Tempe case made the question concrete, because a car packed with sensors was physically capable of detecting an obstacle while the software was not authorised to stop for it. For anyone building or regulating automated driving, the lesson was that the emergency layer cannot be treated as an optional extra that is switched off to keep the ride smooth^[1][2].