1 Robocars & Robotaxis, Steve Raney Utopia or Dystopia? • Minority: central override • Total Recall: interior, Siri – Johnny Cab • WALL-E atrophy. Robocars & Robotaxis GM EN-V, 1:00-3:40, youtube – Connected Car / V2V Google Robotaxi , 15-61s, youtube – Empty movement, smartphone. Talk Outline • Expectations – Provide a feel for design and planning issues – Many conclusions & predictions exist (wrong) – Fertile field for research • Agenda – Quick: Industry status – Quick: Regulation – How it works & what is difficult – Markets & impacts. Industry status • 2014: safer, more pleasant hands/feet-free freeway driving – Lane keeping + traffic jam assist. Advanced Driver Assistance System (ADAS) – Detect driver alertness / ability to take control • Versus robo: robot is responsible for your safety. Freeway + surface. Who ADAS Robo: LIDAR? Robo: When? Google Yes, 300K mi 2016-2018 BMW Yes, 20K mi 2020 Nissan ? 2020 Mercedes 2014 Concept projects Toyota Yes VW/Audi Demo, “by 2019” Academic: Stanford GM Super Cruise, 2015 EN-V: V2V, academic 2020 Ford 2017-2025 Volvo 2014 Continental Yes Radars, 10K mi Regulation vs. Invisible Hand • CA SB 1928 How it works – LIDAR-centric, no V2V
5
Embed
Robocars & Robotaxis, Steve Raney Robocars & Robotaxis · 1 Robocars & Robotaxis, Steve Raney Utopia or Dystopia? • Minority: central override • Total Recall: interior, Siri –
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Robocars & Robotaxis, Steve RaneyUtopia or Dystopia?
• Minority: central override
• Total Recall: interior, Siri– Johnny Cab
• WALL-E atrophy.
Robocars & Robotaxis
GM EN-V, 1:00-3:40, youtube
– Connected Car / V2V
Google Robotaxi, 15-61s, youtube
– Empty movement, smartphone.
Talk Outline
• Expectations
– Provide a feel for design and planning issues
– Many conclusions & predictions exist (wrong)
– Fertile field for research
• Agenda
– Quick: Industry status
– Quick: Regulation
– How it works & what is difficult
– Markets & impacts.
Industry status• 2014: safer, more pleasant hands/feet-free freeway driving
– Lane keeping + traffic jam assist. Advanced Driver Assistance System (ADAS)
– Detect driver alertness / ability to take control
• Versus robo: robot is responsible for your safety. Freeway + surface.
Who ADAS Robo: LIDAR? Robo: When?
Google Yes, 300K mi 2016-2018
BMW Yes, 20K mi 2020
Nissan ? 2020
Mercedes 2014 Concept projects
Toyota Yes
VW/Audi Demo, “by 2019” Academic: Stanford
GM Super Cruise, 2015 EN-V: V2V, academic 2020
Ford 2017-2025
Volvo 2014
Continental Yes Radars, 10K mi
Regulation vs. Invisible Hand
• CA SB 1928
How it works – LIDAR-centric, no V2V
2
Taking a ride in the Google robocar
• You are in the back seat
• Safety driver in driver seat
– Can switch from robot to human/manual driving
• Big red button in-between seats
• Engineer rides shotgun with laptop
– Watches objects being perceived
– Adds notes
• Can later analyze sensor data stream for debugging.
How it works: 3 key tasks
• 1) Know precise location (“localization”)
– Pattern match pre-mapped sensor data stream
• 2) Detect hazards
– Form object trajectory hypotheses
– Identify & avoid danger
• 3) Path planning
• Racing example, 24-49s, youtube
What’s hard about human driving? � Use cases
• Robotaxi product solution? Route, I2V.
Difficult human driving
• Bikes: fast & random
• Erratic drivers
• SF hills
• Surface St. vs. freeway.
Difficult human driving Add robot driving difficulties• Maneuvers that depend on eye contact
– Traffic cops?
– Traffic cop solution? Manual, control room, read gestures
• Reading human intentions (or lack thereof)
– How could robot communicate intentions?
3
Steve Shladover’s Cautionary Insights
• UC PATH expert
• Human errors can be eliminated, but
• Cascading system dependencies introduce new vulnerabilities