This document discusses the benefits of autonomous vehicles and the progress being made towards developing them. It outlines how self-driving cars could end traffic deaths, reduce wasted time and costs from driving, eliminate parking issues, and have environmental benefits. The technology for computer-assisted driving already exists, and fully driverless vehicles may be possible within a decade. Challenges include legal issues, perception capabilities, costs, and completing autonomous algorithms.
2. Wish List
• End wasted driving time
• Greatly reduced traffic fatalities
• End parking hell
• Conduit for much cheaper personal and package
transportation
• Conduit for switch to all electric transportation
• Vehicles that drive themselves
• Pooled by use vehicles
• Drives itself (safely and efficiently)
– You can spend the travel time more productively or not
– No parking
– Comes to you
3. A Few Reasons Why
• Driving yourself is very dangerous..
– 35,000 deaths and over 1,000,000 injuries/yr US
– $230 billion/yr in accident costs (over 2% US GDP)
• Is very costly..
– 5 billion hours (est $1 trillion value) of people’s time (8% of US
GDP)
• Geopolitical, environmental
– 50 billion gallon of imported gasoline
• Replace with 5 billion gallon equivalents produced domestically
• Eliminates 12-15% of US CO2 pollution
• ~600 million urban parking places, up to 10% of urban land
in many cities
• Multiply by 10 for rough world wide impact
4.
5. Bits already on the street
• Computerized steering and braking
• Automatic parallel parking
• Active relative distance cruise control
• Lane drift warnings
• Built in lidar, radar, cameras
• Crash avoidance
• Real time traffic data incorporation
• 15 years of (mostly) self driving cars
– VaMP 1994 – heavy city driving with little human intervention
– Mercedes 1995 – 1600km on autoban, 95% computer only
6. Coming Helpful Tech (near)
• Crash avoidance
– 80% of crashes are from inattention
– Warn today, soon active crash avoidance
– Way out of some legal roadblocks
– The cars talk with one another in local mesh grid
• Navigation improvements
– System incorporates weather, road conditions, traffic
reports, sensors along route, flow of local, google
maps, etc for most efficient route
– Networking with other vehicles for best over all traffic
network flow
7. More Helpful Tech
• Delivery bots
– Military uses these, including all terrain models like Big
Dog
– Imagine them instead of UPS trucks and other human
driven vans. Agile fleets with smaller loads
• Other “sense”
– Tied into traffic lights and other systems.
• Light is always green if there is no relevant cross traffic
• Can know the timing of the others ahead also
– 30% better mileage in city
– Tied to your “cloud” via 3/4G or other means
– Tied to entire internet
8. More near term tech
• Self-parking valet service
– Parking in human driver impossible spaces
– Instead of parking could rent itself out when you
don’t need it
• Mini-car robotic personal transport grids
– Non-personal units
– Automatic routing by wire for first units
9. Medium range tech
• Autopilot
– Human assisted
• Urban delivery bot
– By wire at first
– Then free roving light slow moving bots
• Whistle car
– Comes on demand
– Moves more slowly and perhaps in special lanes
• Robotic Convoy
– Drafting gives up to 45% mileage boost
– Automatic convoys of vehicles doing follow the leader
– Driver or vehicle joins and leaves convoy freely
10. Longer Term Tech
• School of fish test
– Successful avoidance of determined intention to
crash]
• Better information, much faster reaction times
– Need fish like cooperative movements
• Learning from human drivers
– Autopilot mode watches and learns even when
often
• Multiple units share and correlate data
• Test suite generation
11. When
• Moore’s law gives 256 fold increase in
computational power this decade
• While we need algorithmic improvements there
are no large conceptual missing parts
• No big money invested yet but the payoff
potential is larger than the Manhattan project
• Blockers are
– Legal
– Perception
– Current cost
– Some algorithms
12. How Expensive?
• Today’s fully autonomous cares are very expensive
– Moore’s law is our friend
– Much better algorithms including through simulations
• More cheaper specialized cars
• How expensive are ordinary cars
– $240/mo + 1,000 miles gas + 40/mo insurance ~= $280/mo
• Robocar (on demand) ~= 1000 miles * per mile cost
– For muture tech this can be as low as $0.07 / mi = $70, and
$0.01 for light urban vehicles
• Savings on materials