PoliMove's driverless car had 1.5 seconds to dodge Unimore's safety stop at Imola

By its team's account, Eva saw it coming. PoliMove says its car detected Unimore's car, Gianna, ahead in Imola's Rivazza corners, identified the danger and began an evasive maneuver ^[6]. It hit Gianna anyway ^[2]. At 155 mph, the speed in Wired's headline, a car covers about 341 feet in 1.5 seconds, which was Eva's gap when Gianna braked ^[1][5][1].

When one car runs into the back of another, people usually conclude the car behind reacted badly. PoliMove's account splits the reaction into parts. Perception, decision-making, trajectory replanning, actuator response and vehicle dynamics each add delay, the team said, and once Gianna braked that severely mid-corner, a collision was physically impossible to avoid ^[7]. Wired concludes that faster processors alone cannot fix this, because a car can recognize a hazard and still be unable to respond before physics takes over ^[14]. A2RL's head of sporting, Alexander Winkler, had pointed to this part of the track before the final. "The car will be blind for a certain point because the track drops off," he said, adding that a car behind a stopped vehicle could have less than a second to react ^[12].

Autonomous racing is pitched as a laboratory, a way to test perception, prediction and vehicle control at extreme speed without putting a driver at risk ^[15]. At Imola, the first thing to fail was the fallback. Unimore said Gianna started a safety stop after losing all data from its lidar and radar, since GPS alone could not locate the car accurately enough to keep racing ^[3]. The team said future systems could use cameras and an inertial measurement unit to continue at a slower pace until precise localization returned ^[4]. The car behind, 1.5 seconds back, met the hard stop Unimore's software chose: about 1 g of braking in the middle of a corner ^[5][7].

Three of the five cars in the final did not finish ^[2]. Only one of those three had to react to another car. TUM, a two-time winner, withdrew its car, Hailey, after the rear-left brake stuck at low-to-medium pressure during the formation lap ^[10]. Gianna stopped on sensor loss ^[3]. Eva was the car that had to react, and PoliMove still placed third, with a replacement car standing in at the podium ceremony ^[9]. Every cause here comes from the teams' own explanations to Wired ^[3][6][10]. The published account does not include telemetry.

Preparation time was short. Teams got nine days of physical testing, in rain and hail, for the league's first international race away from Abu Dhabi's Yas Marina Circuit ^[11]. Nicola Palarchi, engineering director at Aspire, the company that founded A2RL, defended the choice of track. "Because everybody can do 'easy', right?" he said. "We have to show we go where it matters." ^[13]

A team building on an autonomous stack can map Imola onto a two-by-two. One axis is what the vehicle does when a primary sensor drops out: stop hard, or keep moving slowly on whatever it still trusts. The other axis is whether anything is following it closely at speed. With nothing behind, a hard stop is the safe choice and a slow degraded mode costs only time. With a close follower, the slow mode is the option Unimore described for its future systems ^[4]. The hard stop is what happened at Rivazza, where about 1 g of braking left the car behind 1.5 seconds ^[5]. Which box a fallback lands in depends on the follower's total delay from sensor to brakes, and PoliMove's five sources of delay are a usable list for measuring it ^[7].