AUTONOMY AND SMART URBAN MOBILITY November 15, 2017 Emilio Frazzoli Professor of Dynamic Systems and Control, ETH Zürich Co-Founder and CTO
Why Self-driving Vehicles?
A financial perspective on personal mobility (US Market) Safety: Cost of a statistical life : $9.1M 2014 NHTSA report: - Economic cost of road accidents: ~ $277B/year. - Societal harm of road accidents: ~ $594B/year Cost of congestion: Texas Transportation Institute, 2012: ~ $100B/year Health costs of congestion: Harvard School of Public Health, 2010: ~ $50B/year Increased productivity/leisure: Estimate $1.2T/year Car sharing: Assuming a sharing factor of 4, estimate $1.8T/year of benefits to individuals. Other studies [Burns et al., 13, Fagnant, Kockelman 14] suggest higher sharing factors, up to ~10. Sharing Productivity Safety (Economic Cost) Safety (Societal Harm) Congestion Health
(Autonomous) Mobility on Demand in Singapore Time of day Red dot: empty car Blue dot: passenger (in car/in queue) - Optimal algorithms for fleet management, - Minimize operational cost with quality of service constraints
5
IMPACT OF AUTONOMOUS URBAN TRANSPORTATION IN SINGAPORE VIDEO ~800,000 passenger vehicles! ~300,000 shared autonomous vehicles 6 K. Spieser, K. Treleaven, R. Zhang, E. Frazzoli, D. Morton, and M. Pavone. Toward a systematic approach to the design and evaluation of automated MoD systems: a case study in Singapore. In S. Beiker, editor, Road Vehicle Automation, Lecture Notes in Mobility. Springer, 2014.
HOW MUCH WOULD IT COST? 100% Notes Unit Margin (17%) Unit Margin (46%) Increase margin supports fixed operating cost and profit 75% Insurance (7%) Fuel (13%) Insurance cost lower due to safer driving, strong documentation for events that do occur 50% Main. (5%) Vehicle (8%) Insurance(4%) Fuel (10%) Main. (7%) Fuel cost decreased due to usage of electric cars and more efficient operation Maintenance cost increased due to higher safety standards, additional cost of maintaining autonomous hardware 25% Labor (50%) Vehicle (23%) Vehicle cost higher due to sensors, newer cars and assumed shorter life (conservative) 0% Current taxi Labor (10%) Estimated AV taxi Labor cost reduced but not eliminated. Assume teleoperation supervision (10 cars to one operator) and cleaning / monitoring of vehicles 7 Source: Various studies of taxi cost profile
IMPACT ON THE LABOR MARKET Traditional On-Demand Mobility requires one out of 7 people to be a driver - Impact mainly on increased supply of mobility rather than job loss - Cost reduction may be an issue, may be balanced by added value in service 8
WHEN WILL AUTONOMOUS VEHICLES ARRIVE? Level 4/5 Autonomous Vehicle Penetration Mobility fleets 100% 80% 60% 2030 25% of US miles by autonomous vehicles 3 2040 75% of all vehicles are autonomous 4 2050 Traffic fatalities reduced by 90% 5 2060 Cities restrict human driving 2 Personally owned vehicles 40% 20% 0% 2017 Pilots in: Singapore, Boston, Phoenix, San Francisco, Pittsburgh 2025 100 cities with large AV fleets 2 2025-2030 Level 4: Highway 1 Slow increase as individuals purchase new self-driving cars Legend Personally owned vehicles Mobility fleets 2010 2020 2030 2040 2050 2060 9 Sources: 1 Nissan, Toyota, Ford, BMW etc.; 2 nutonomy estimates; 3 BCG; 4 IEEE; 5 McKinsey
10 CURRENT STATUS: A TYPICAL DAY IN SINGAPORE
OUR MISSION To radically improve the safety, efficiency, and accessibility of transportation in cities worldwide 12
Realizing the value of self-driving cars Safety: Severe human-factors limitations to safety of systems with humans in/on the loop. it may be impossible to demonstrate the reliability of high-performing autonomous vehicles (i.e., ones with failure rates comparable to or better than human failure rates) to any reasonable degree of precision. RAND Co. 2016 Time value of driving Needs Level 4: value mostly to individuals Car sharing Needs Level 4: societal impact in addition to individual convenience Level 4 automation ( Full Autonomy ) is essential to capture the benefits of self-driving cars The game-changing feature of AVs is their ability to drive themselves with nobody on board.
Paths to Level 4 Level of Automation 4-3 - 2-1 - Today ~2030 Today Prototype/closed course Fleet, limited geography Scale / Scope Mass deployment, everywhere
Product vs. Service AVs as a consumer product AVs as a service (MaaS) Scope Where and when the AV capabilities must function Financials Cost constraints Infrastructure Maps, dealers, service Everywhere, all the time Comparable to the cost of the vehicle and/or driver s time. PV of the driver s time: ~23,000 USD for a 10-year lifetime Global scale, immediately Geo-, time-, weather-fenced operation Comparable to the cost of hiring a driver (crew) > 100,000 USD per year Scale (sub)linearly with the user base Servicing and Maintenance Most high-tech sensors etc. not user serviceable yet Servicing/maintenance crews already on roster.