Fast Forwarding to a World of On-Demand Air Transportation

Transcription

1 Fast Forwarding to a World of On-Demand Air Transportation Smart Urban Transportation Forum IMA University of Minnesota May 15, 2017 Jon Petersen Head of Data Science, Advanced Programs Uber Technologies

2 dynamic pricing spatio-temporal forecasting dispatch Rich problems revenue management driver positioning latent demand modeling uberpool stochastic routing anomaly detection ubereats map matching many more

3

4 Uber Elevate

5 Noise Speed Efficiency Emissions Safety (thus cost)

6 Over a dozen VTOL companies developing prototypes to deploy within the next three years Volocopter Lilium Zee.Aero Ehang Joby Aviation A^3/Airbus

7 Distributed Electric Propulsion + Fixed Wing Very quiet compared with combustion engines, Many rotors -> lower tip speed Inherently safer: Redundancy is tolerant of failures, partial autonomy 2X safer than driving Emissions: Electric has zero operational emissions 10x more efficient: 3-4x improvement from tiltrotor, 3x improvement in fixed wing

8

9 Vehicle Designs multicopter autogyro helicopter co-axial helicopter compound helicopter lift + cruise vectored thrust tilt duct tilt wing tilt rotor en-route cruise speed [mph] lift/drag disc-loading [lbs/ft 2 ] efficient cruise FOM efficient takeoff & landing empty weight/ gross weight en-route energy [kwh/mile]

10

11

12

13 Flux Optimizer Beyond the Vehicle Where to build infrastructure? What are the network dynamics? What does a good network need?

14 Flux Optimizer Under the Hood 1. Estimate Demand 2. Generate Candidate Hubs 3. Hub Optimization 4. VTOL Route Optimization

15 Flux Optimizer Under the Hood 1. Estimate Demand 2. Generate Candidate Hubs 3. Hub Optimization k-means clustering on trip points 4. VTOL Route Optimization centroids define candidate hubs also supports hubs defined by the user

16 Flux Optimizer Under the Hood GOAL maximize time savings of all riders 1. Estimate Demand 2. Generate Candidate Hubs 3. Hub Optimization An itinerary (o,d,i,j) is valid if the time saved s(o,d,i,j) exceeds some threshold s*: 2 s(o, d, i, j) = x(o, d) {z } ground time s x(o, d) 4 (o, i)+ (L)+ (i, j)++ (U)+ (j, d) 5 {z } VTOL time 3 4. VTOL Route Optimization

17 Flux Optimizer Under the Hood GIVEN all eligible VTOL itineraries set of candidate hubs H user inputs 1. Estimate Demand Optimize subset of hubs H* to maximize coverage 2. Generate Candidate Hubs 3. Hub Optimization 4. VTOL Route Optimization

18 Flux Optimizer Under the Hood 1. Estimate Demand GIVEN optimal hub placements trips DECIDE sequence of flight legs assign riders to VTOLs when/where/how much to recharge when/where to deadhead when to depart (savings vs. load) 2. Generate Candidate Hubs 3. Hub Optimization 4. VTOL Route Optimization

19 Hub Optimization

20 DFW Hub Optimization

21 DFW Rolling Out Infrastructure 1, 2 Rider Coverage = 15.2% + 3, 4, 5 + 6, 7, 8, 9, 10 Rider Coverage = 42.2% Rider Coverage = 47.3%

22 Dubai Hub Optimization 2 Hubs 5 Hubs 10 Hubs 50 Hubs 23% 52% 69% 86% % Trip Coverage

23 Minneapolis/St. Paul Hub Optimization

24 VTOL Route Optimization

25 VTOL Routing Model Multicommodity Network Flow Model

26 VTOL Routing Model ground arcs (battery charging) Better modeling approach: identify path P that is a minimum cover valid for (v,p) then lift following arcs following the final arc in P

27 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec DECEND 45 sec RECHARGE MINUTES [20-90%] TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

28 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec DECEND 45 sec RECHARGE MINUTES [20-90%] TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

29 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] Demand may be met using OUTPUT evtol RANGE mi ~ 40% of vehicle capacity EN-ROUTE SPEED [mph] ASCEND 75 sec RECHARGE MINUTES [20-90%] Load factor DECEND ~ 1.55 per flight 45 sec TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

30 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec DECEND 45 sec RECHARGE MINUTES [20-90%] TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

31 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUMM GROUND TIME [min] most sharing of VTOLs is attained with 3 seats routes with highest load factors Frisco DFW Downtown Dallas OU UT TPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec DECEND 45 sec RECHARGE MINUTES [20-90%] TRIP VOLUME [1.0 = reference] MIN 10 VTOL RANGE [m] 20 MIN VTOL RANGE [m] 10 20

32 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec DECEND 45 sec RECHARGE MINUTES [20-90%] TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

33 e V TO L CO UNT [ % CA PA CI T Y ] North Denton # O F HUB S E AT CO UNT AIRSPACE S E P. MI NI MUM G RO UND T I ME [m i n ] 5 min MGT! 2 mi evtol RA NGE mi 3 min MGT E N-RO UT E S P E E D [ mp h ] ASCEN D 75 sec RE CHA RG E MI NUT E S [ % ] DEC END 45 sec T RI P V O L UME [ 1. 0 = re f e re n c e ] MI N V TO L RA NG E [ m] MI N V TO L RA NG E [ m] OUTPUT - Las Colinas INPUT ASSUMPTIONS DAL LA S, T X

34 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec DECEND 45 sec RECHARGE MINUTES [20-90%] TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

35 1705 mph cruise AIRSPACE S E P.! 2 mi evtol RA NGE mi e V TO L CO UNT [ % CA PA CI T Y ] S E AT CO UNT mph cruise MI NI MUM G RO UND T I ME [m i n ] E N-RO UT E S P E E D [m p h ] ASCEN D 75 sec RE CHA RG E MI NUT E S [ % ] DEC END 45 sec T RI P V O L UME [ 1. 0 = re f e re n c e ] MI N V TO L RA NG E [ m] MI N V TO L RA NG E [ m] OUTPUT # O F HUB INPUT ASSUMPTIONS DAL LA S, T X

36 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec RECHARGE MINUTES [20-90%] DECEND 45 sec TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

37 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE ACE SEP.! 2 mi NPUT I evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] Charging only at mega vertiports OUTPUT RIDERS SERVED - 4.0% TIME SAVED + 0.0% evtol RANGE mi ASCEND 75 sec DECENDD 45 sec EN-ROUTE SPEED [mph] RECHARGE MINUTES [20-90%] TRIP VOLUME [1.0 = reference] MIN VTOL RANGE [m] z BATTERY CHARGE AT LANDING -17% evtol UTILIZATION - 2.3% DEADHEADS PER evtol +0.5% MIN SAVINGS THRESHOLD [m] MIN VTOL RANGE [m] LOAD FACTOR +1.5%

38 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec RECHARGE MINUTES [20-90%] DECEND 45 sec TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

39 ASSUMPTIONS DALLAS, TX # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec RECHARGE MINUTES [20-90%] DECEND 45 sec TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

40 ASSUMPTIONS DUBAI, UAE # OF HUB 5 AIRSPACE SEP.! 2 mi INPUT evtol COUNT [% CAPACITY] SEAT COUNT MINIMUM GROUND TIME [min] OUTPUT evtol RANGE mi EN-ROUTE SPEED [mph] ASCEND 75 sec RECHARGE MINUTES [20-90%] DECEND 45 sec TRIP VOLUME [1.0 = reference] MIN SAVINGS THRESHOLD [%] MIN VTOL RANGE [m] 10 20

41 DUBAI, UAE # VTOLs Rider Capacity Minimum Ground Time Cruise Speed Trip Volume (Rider Savings) Min Rider Savings

42 What if pick-up times were longer? Simulate ETAs +10%, +25%, +50%

43 Ongoing & Future Work Incorporating stochasticity vehicle design (speed, battery consumption) VTOL travel times latent demand Air Traffic Management Flight planning for noise abatement Unmanned Aircraft Systems Traffic Management (UTM) Irregular Operations Marketplace Optimization: Pricing, Dispatch, Pooling

44 Thanks! Questions? more about Uber Elevate: Uber Elevate summit videos: Questions, comments: