Uber Will Rent Scooters Through Its App in Partnership with Lime • Hacker News Board

Re: Uber Will Rent Scooters Through Its App in Partnership with Lime

#181

Post by CamperBob2 » Thu, Jul 12, 2018, 3:22 AM UTC

maym86 wrote: post #161

CamperBob2 wrote: post #95

Earlier quoted context omitted.

They are going to self-drive everywhere, I suspect. It's much easier to make a scooter or a bicycle autonomous than a car. There's at least one company around here working on self-delivering bicycles. The idea is that you call one up on your phone and wait by the curb, and the bike will come cruising up to you. When you're done, you just hop off, and the bike cruises to its next renter (or charger) rather than lying…

> It's much easier to make a scooter or a bicycle autonomous than a car. No it's not. Less stable platform. Exact same issues with obstacle detection and avoiance, path planning and scene understanding with less available power for sensors and compute. Higher relative sensor cost. Fewer closed environments like highways. Added issue of being able to steal them more easily.

None of those factors are of any importance whatsoever next to the key one: less likely to kill people when something goes wrong.

Stealing a GPS-tagged autonomous bike or scooter is also not going to go especially well for whoever tries it.

Re: Uber Will Rent Scooters Through Its App in Partnership with Lime

#182

Post by boxcardavin » Thu, Jul 12, 2018, 8:53 PM UTC

maym86 wrote: post #161

CamperBob2 wrote: post #95

Earlier quoted context omitted.

They are going to self-drive everywhere, I suspect. It's much easier to make a scooter or a bicycle autonomous than a car. There's at least one company around here working on self-delivering bicycles. The idea is that you call one up on your phone and wait by the curb, and the bike will come cruising up to you. When you're done, you just hop off, and the bike cruises to its next renter (or charger) rather than lying…

> It's much easier to make a scooter or a bicycle autonomous than a car. No it's not. Less stable platform. Exact same issues with obstacle detection and avoiance, path planning and scene understanding with less available power for sensors and compute. Higher relative sensor cost. Fewer closed environments like highways. Added issue of being able to steal them more easily.

Yo, I'm with that company doing the AV scooters and bikes, and yes it's much easier to nav than a car, we're looking at two magnitudes easier as far as computing power. We won't do autonomous with riders on-board so we can go slow and be conservative. Sensors are more than two orders of magnitude cheaper, we can use ToF laser rangefinders that are close to mobile phone grade.

Re: Uber Will Rent Scooters Through Its App in Partnership with Lime

#183

Post by boxcardavin » Thu, Jul 12, 2018, 8:56 PM UTC

CamperBob2 wrote: post #181

maym86 wrote: post #161

Earlier quoted context omitted.

> It's much easier to make a scooter or a bicycle autonomous than a car. No it's not. Less stable platform. Exact same issues with obstacle detection and avoiance, path planning and scene understanding with less available power for sensors and compute. Higher relative sensor cost. Fewer closed environments like highways. Added issue of being able to steal them more easily.

None of those factors are of any importance whatsoever next to the key one: less likely to kill people when something goes wrong. Stealing a GPS-tagged autonomous bike or scooter is also not going to go especially well for whoever tries it.

Exactly, our mistakes are not inherently catastrophic. Stability is actually much harder than navigating.

Re: Uber Will Rent Scooters Through Its App in Partnership with Lime

#184

Post by maym86 » Thu, Jul 12, 2018, 10:08 PM UTC

boxcardavin wrote: post #182

maym86 wrote: post #161

Earlier quoted context omitted.

> It's much easier to make a scooter or a bicycle autonomous than a car. No it's not. Less stable platform. Exact same issues with obstacle detection and avoiance, path planning and scene understanding with less available power for sensors and compute. Higher relative sensor cost. Fewer closed environments like highways. Added issue of being able to steal them more easily.

Yo, I'm with that company doing the AV scooters and bikes, and yes it's much easier to nav than a car, we're looking at two magnitudes easier as far as computing power. We won't do autonomous with riders on-board so we can go slow and be conservative. Sensors are more than two orders of magnitude cheaper, we can use ToF laser rangefinders that are close to mobile phone grade.

Where are you testing?