First autonomous accident…

Yesterday we had our first (and hopefully our last!) small accident with our autonomous vehicles. We were just at the end of a highly successful driving session (165Km): the guys in charge of the autonomous vehicle (the follower) stopped the vehicle just behind the leader, and got off the vehicle. They disengaged the perception system, but unfortunately they forgot to disengage the x-by-wire and planning systems (well, both of them thought it was the other guy who disengaged everything and got off to relax).

Suddenly the vehicle started moving towards the vehicle in front (with no perception, only GPS waypoints were used) and the follower hit the leader. Luckily it was a slow speed hit and everybody was away since the vehicles were parked, but the chassis got some little damages. The frontal sensors were protected by the bull bar, so no damage and no miscalibration occurred, but now the leader has a small dent in the back part. We don’t know whether to fix it or not: after all this is part of history anyway!

Well, in these cases, there’s no e-stop or other mechanism that can avoid this, since everybody thought everything was disengaged and were relaxing away form the vehicles. This showed a design problem: we should have a very visible device (like a light or something like that) which is on whenever the autonomous driver is engaged.

6 Comments

  1. roberto says:

    Well, good to know that nothing happened to people and equipments, a small visible damage on one of the cars in nothing compared to 1000 (or more) km already performed.
    Hope you quickly solve the classical bureaucratic problems at the Russian border.
    Go ahead.

  2. dom says:

    Human error, then. One day, when the roads are full of autonomous vehicles, we can say goodbye to human error, and then we won’t need to design vehicles that are 90% devoted to crash safety, and accidents like this will not even leave a dent…

  3. Shurhenchik says:

    On my car it is a lot of dents, and I assure you that it does not influence speed of movement! ;) ))))))))))))))))

  4. yaguar says:

    Just a thought: the handbrake is not annuled yet :)

  5. Ermanno says:

    “… This showed a design problem: we should have a very visible device (like a light or something like that) which is on whenever the autonomous driver is engaged.”
    I do not agree with that; the design problem, IMHO, is that both x-by-wire and planning systems should not be allowed to act whether the perception system is not engaged (with a metaphor, you don’t run with your eyes shut).
    Of course, I think it will be fixed when all the various systems will be duly integrated; in fact, now you are just testing and experimenting.
    Last (following previous comment by Dom), the real “human error” seems the quantity of problems the authorities are raising to you at the customs checks… :)
    Cheer up! You will succeed.

  6. [...] The trip took them 16,000 kilometers through nine countries in 100 days. In Russia (where the car had its first accident, due to human error), the team nearly notched another record: The first autonomous vehicle to be [...]