Today’s Headlines

Get state headlines at Streetsblog CA, national headlines at Streetsblog USA

  • murphstahoe

    The sturm and drang over the AV stuff is amusing, but it won’t really drop the momentum. A lot of the testing can be done by taking a car out with cameras, then piping the footage into the AV and do it virtually.

    Here was a comparison I told to a friend.

    When I was 17, I was in a car crash. I don’t really recall much about it because I was concussed – but it was definitely my fault. I was making a left turn into a pizza place, and either I turned left in front of oncoming traffic or as I moved to the middle lane my car slid on the ice.

    As I analyzed it later, I could see some of the possibilities of how I caused the crash and took away some lessons, however inaccurate.

    In the Uber crash, they have camera footage and other data from the car so they can tell exactly what happened. The engineers can figure out exactly what happened, and if the accident could have been avoided by the car, they can use the data from the crash to improve the programming such that this situation would be avoided, and actually test against the real footage from the crash to evaluate their improvements.

    So in a human crash, one human will take away some fuzzy learning. In the AV crash, detailed analysis will inform substantial learning that will inform changes that EVERY AV will instantly utilize.

    I’ll take the latter.

  • thielges

    It would be great if AV companies shared crash and safety data to accelerate learning but I have a hunch that most companies will keep a tight grip on anything that might reveal their technical secrets. More likely they release only what they are legally required to release.
    Like human drivers, AVs are constantly faced with making a tradeoff between safety and speed. Employees of AV companies then are faced with how to tweak the parameters of their product to please their customers by quickly reaching destinations while not moving so fast as to break things (or people). It is likely that the Uber’s sensors clearly detected something moving across the lanes of traffic but the higher level perception and control logic places a low probability that the “something” required braking to avoid. It is not clear whether the NTSB has the authority, skills, or motivation to suss that out.
    This tradeoff between safety and speed is why AVs should be held to a higher standard than ‘safer than the average human driver’. At least in the USA the pool of drivers includes a large amount of incompetence, bringing the average way down. That is too low of a bar. If that’s the metric that AVs need to meet then expect that AVs may not necessarily get safer but instead get faster and more aggressive as engineers spend their “safety credit” from improvements in detection and perception on going faster rather than driving safer.

  • murphstahoe

    My hope is that the fact that riding in a car disengaged from the sullen process of actually doing the driving will lead to people being less concerned about the speed. Much like 1 hour on a train is better than 50 minutes in a car.

    One point someone made to me – in the world of self driving fleets of shared cars, you should be able to pull up your destination and the app should be able to tell you exactly when you will arrive at your destination, factoring in time for the car to get to you and time to destination, which will factor in traffic conditions, weather, and the fact that the car won’t speed (in theory) – e.g. you can’t decide you are late so start running red lights or putting the pedal to the floor.

    So if you need to be somewhere at 7 PM, you can tell your phone to schedule a pickup at the time you need in order to arrive on time which will be precise. And you won’t be able to tell your date that you left on time but traffic was a nightmare – you leave late, you arrive late, period.

    Changing this whole paradigm is something I hope will lessen the desire to have the cars be unsafe. It is – what it is.

  • thielges

    I’m sure you’re right for most people. Even today there are plenty of drivers (especially professional drivers) who are chill and drive safe without being aggressive. There might be a market for people who want a competitive edge on the roads and can pay for it. I wonder what would happen if placebo “go faster” buttons were put on conventional transit. Some might press it out of earnest desire to get to their destination quicker.

  • Mario Tanev

    After the shit SFMTA pulled with dockless bike sharing, who can blame the scooter operators for not waiting for SFMTA to establish a process, which then ends up being a monopoly.