highlandsflyer
Retro Wizard
- Feedback
- View
I wonder how this is going to pan out.
They will presumably blame the human..
They will presumably blame the human..
Autopilot Car Kills Driver!
- Thread starter highlandsflyer
- Start date
- Feedback
- View
Re:
From what I gather, it's a freak accident due to extreme circumstances. Neither the car nor the driver could apparently see the truck + trailer that was crossing the road.
https://www.teslamotors.com/blog/tragic-loss
Kinda hard to blame anyone for that.
From what I gather, it's a freak accident due to extreme circumstances. Neither the car nor the driver could apparently see the truck + trailer that was crossing the road.
https://www.teslamotors.com/blog/tragic-loss
Kinda hard to blame anyone for that.
highlandsflyer
Retro Wizard
- Feedback
- View
Re:
Not so. If a human made the error easy to blame. If not, just as easy.
No system that takes control should be forgiven for making such a mistake.
Not so. If a human made the error easy to blame. If not, just as easy.
No system that takes control should be forgiven for making such a mistake.
videojetman
Retrobike Rider
- Feedback
- View
just goes to show technology is flawed. you will never be able to replace the driver completely.
although through this tragedy it will improve this technology.
but very sad a life was lost.
although through this tragedy it will improve this technology.
but very sad a life was lost.
Re:
Ya right.
Just remember, very very soon you will never ever be able to speed or break any road rules, as each vehicle will be transponder equipped and your driving will be real-time computer monitored ... any infractions will be immediately known and recorded.
Shitty deal.
Ya right.
Just remember, very very soon you will never ever be able to speed or break any road rules, as each vehicle will be transponder equipped and your driving will be real-time computer monitored ... any infractions will be immediately known and recorded.
Shitty deal.
Nah, it shows how shoddy Teslas testing is. Most manufacturers wouldn't have a beta test in production, in the wild.videojetman":3idyy5t9 said:
A beta test should be a (very) limited production run under (very) tight control. Especially with safety critical stuff like this. More like 100 cars with weekly/fortnightly data downloads and software updates. Plus driver training so they know what they are getting into.
Not doing it as an update for anyone who wants it.
Not seeing a slab sided truck is a serious dropped bollock. Should have three completely different and essentially independant sets of sensors, radar, sonar, camera, lidar, ultrasound, IR, take your pick. Half of them "see" in black and white anyway.
I hope Tesla get seriously reamed out for this.
- Feedback
- View
Tessla's sensors were too low resulting in blind spots. Their excuse was a white truck against a white background despite ultrasonics/ radar etc.
They say they've done some 163,000,000 miles of testing.
Total driverless is some decades away - cars will need to be almost sentient plus there cant be any communication blank spots with GPS and 5g/ 6g networks. Things like emergency vehicles will need to communicate with nearby vehicles to move to a safe position while it passes, will zebra crossings still be in use? Crossings and traffic lights will need to communicate with driverless vehicles on cross platform networks.
*A warning from the current 'smartmeter' rollout from the electricity supply industry says a lot. The new smart meters being installed have to then be taken out if you change supplier, they are not cross compatible with other companies' software/ comms systems. Now transfer that over to say Ford and PSA for arguments sake - will US congress allow for US intellectual property to be accessed by other companies?
The ingredients are there, just a little more processing power and faster networks.
I look forward to it!
They say they've done some 163,000,000 miles of testing.
Total driverless is some decades away - cars will need to be almost sentient plus there cant be any communication blank spots with GPS and 5g/ 6g networks. Things like emergency vehicles will need to communicate with nearby vehicles to move to a safe position while it passes, will zebra crossings still be in use? Crossings and traffic lights will need to communicate with driverless vehicles on cross platform networks.
*A warning from the current 'smartmeter' rollout from the electricity supply industry says a lot. The new smart meters being installed have to then be taken out if you change supplier, they are not cross compatible with other companies' software/ comms systems. Now transfer that over to say Ford and PSA for arguments sake - will US congress allow for US intellectual property to be accessed by other companies?
The ingredients are there, just a little more processing power and faster networks.
I look forward to it!
- Feedback
- View
Were both vehicles driverless? If so would the accident have happened? I get the bit about the sensors but an isolated vehicle isn't a true test in my view.
- Feedback
- View
Re:
As much as you can make a case for testing on closed roads etc, you can never come up with all the situations that a car faces in real traffic. There is just no substitute for real-life testing. As long as that testing doesn't substantially increase the amount of accidents and the death toll, there shouldn't be any problem with that.
Sensor logs will be looked at and there will probably be reconstructions of the accident (in a closed environment and without any drivers in the cars) to figure out why exactly the car didn't see the truck and how to improve things.
Accidents happen. We can use them to make things better by tweaking the sensors and/or algorithms.
Or we can use this one single accident to delay or ban a technology that already has a much better track record than humans have on their own.
As much as you can make a case for testing on closed roads etc, you can never come up with all the situations that a car faces in real traffic. There is just no substitute for real-life testing. As long as that testing doesn't substantially increase the amount of accidents and the death toll, there shouldn't be any problem with that.
Sensor logs will be looked at and there will probably be reconstructions of the accident (in a closed environment and without any drivers in the cars) to figure out why exactly the car didn't see the truck and how to improve things.
Accidents happen. We can use them to make things better by tweaking the sensors and/or algorithms.
Or we can use this one single accident to delay or ban a technology that already has a much better track record than humans have on their own.
