motion planning for multiple autonomous vehicles: chapter 7a - semi autonomous its
DESCRIPTION
This series of presentations cover my thesis titled "Motion Planning for Multiple Autonomous Vehicles". The presentations are intended for general audience without much prior knowledge of the subject, and not specifically focused upon experts of the field. The thesis website contains links to table of contents, complete text, videos, presentations and other things; available at: http://rkala.in/autonomousvehiclesvideos.htmlTRANSCRIPT
School of Systems, Engineering, University of Reading
rkala.99k.orgApril, 2013
Motion Planning for Multiple Autonomous Vehicles
Rahul Kala
Semi-Autonomous Intelligent Transportation SystemPresentation of paper: R. Kala, K. Warwick (2015) Intelligent
Transportation System with Diverse Semi-Autonomous Vehicles, International Journal of Computational Intelligent Systems, 8(5): 886-899.
Motion Planning for Multiple Autonomous Vehicles
Key Contributions• The approach presents an integrated study of
an intelligent transportation system covering all the various concepts which are separately studied in the literature.
• The study proposes architecture of the transportation systems of the future covering both decentralized vehicle control and a centralized management control.
• The approach is designed for diverse semi-autonomous vehicles operating in a scalable environment, which is the likely future of the transportation system.
• The approach is a positive step towards creation of a traffic simulation tool for diverse and unorganized traffic.
rkala.99k.org
Motion Planning for Multiple Autonomous Vehicles
Assumption• All semi-autonomous vehicles, or all can
communicate• All vehicles can be tracked• There might still be some human driven
vehicles• Vehicles have very diverse speeds
Key idea• Explore all the possibilities with such an
assumption• Enable vehicles cooperatively reach their
destination in the best way• Make transportation rules dynamic
rkala.99k.org
Motion Planning for Multiple Autonomous Vehicles
Proposed Intelligent Transportation System Architecture
rkala.99k.org
Central Information System
Vehicle Route Planning
Vehicle Control
Vehicle MonitoringTraffic Signal
Module
Speed Lane Module
Vehicle Motion Planning
Lane Booking Road
Booking
Scenario Specification
Map/Initial conditions
Traffic at roads
Speed of vehicles at lanes
Position/ Speed
Speed Limit
Booking Specifications
Signal state
Traffic Info.
Lane change, Follow, Stop/Start, Turn
Booked?Booked?
Motion Planning for Multiple Autonomous Vehicles
Traffic Lightning System • Since all vehicles are tracked, lights can change as soon
as all vehicles at a particular side are clear
rkala.99k.org
• Minimize maximum waiting timeAim:
• Change lights to allow the most waiting vehicle (and all other vehicles in that side) to pass by.Concept 1
• A vehicle may not simply drive through without waiting for a light change (if some other vehicle at some other side is waiting)
Concept 2
• No vehicle left in the current side, threshold number of vehicles have passed by, threshold time has passed by.
Change of light happens if
Motion Planning for Multiple Autonomous Vehicles
Traffic Lighting System
rkala.99k.org
1
2
35
4 6
1
2 3
Order of change of traffic lights. Numbers denote the order of appearance in the crossing scenario for the
present vehicles.
Motion Planning for Multiple Autonomous Vehicles
Traffic Lighting System
rkala.99k.org
Comparisons
Proposed System
Cyclic Light
Changes
Light change
after time threshold (current system)
Motion Planning for Multiple Autonomous Vehicles
Traffic Lighting System
rkala.99k.org
3 135 267 399 531 663 795 927 10591191132314550
100
200
300
400
500
600
700
Number of Vehicles
Ave
rage
tim
e of
tra
vers
al
Varying traffic density
Motion Planning for Multiple Autonomous Vehicles
Traffic Lighting System
rkala.99k.org
3 147 291 435 579 723 867 10111155129914430
100
200
300
400
500
600
Number of Vehicles
Ave
rage
Tim
e of
Tra
vers
al
Varying traffic density with traffic from one side blocked
Motion Planning for Multiple Autonomous Vehicles
Traffic Lighting System
rkala.99k.org
1 6 11 16 21 26 31 36 41 46 51 56 610
100
200
300
400
500
600
700
800
900
Proposed System
System with definite time of change
System with cyclic changes
time_th
Ave
rage
tim
e of
tra
vers
al
Varying maximum time per light change for dense traffic
Motion Planning for Multiple Autonomous Vehicles
Traffic Lighting System
rkala.99k.org
Varying maximum time per light change for light traffic
1 29 57 85 1131411691972252532813093373650
100
200
300
400
500
600
time_th
Ave
rage
Tra
vers
al T
ime
Motion Planning for Multiple Autonomous Vehicles
Traffic Lightning System - Design Choice
rkala.99k.org
Make frequent light changes for dense traffic (towards first come first serve system, limits maximum waiting time)
Make infrequent light changes for dense traffic (smaller average traversal time, smaller time wasted in transition between changes)
Motion Planning for Multiple Autonomous Vehicles
Speed Lane• Since the vehicles are tracked and under
communication, speed distribution between lanes can be made dynamic
• Assumption: Speeds uniformly distributed in the speed band (between slowest and fastest current vehicle on the road)
• Concept 1: Divide the speed band by weights and distribute between the lanes
• Concept 2: Higher speed vehicle can jump to lower speed lane (for overtaking) but not vice versa rkala.99k.org
Motion Planning for Multiple Autonomous Vehicles
Speed Lanes
rkala.99k.org
Lowest speed capability vehicle
Highest speed capability vehicle
Speed Band(all other vehicles assumed to be uniformly distributed in this band)
Speeds
Speed limit of individual
lanes
Weighted speed division
Motion Planning for Multiple Autonomous Vehicles
Speed Lanes
rkala.99k.org
0
100
200
300
400
500
600
700
With speed lanesWithout speed lanes
w
Ave
rage
tra
vers
al t
ime
Comparisons for densely occupied scenario
Comparisons with a system with no speed lane – any vehicle can go anywhere
Motion Planning for Multiple Autonomous Vehicles
Speed Lanes
rkala.99k.orgComparisons for lightly occupied scenario
Speed lanes are a bad idea when traffic density is low, all diverse vehicles having access to both lanes can quickly criss-cross and overtake, not having to follow a vehicle as the other lane is reserved for high speed vehicles only
0
100
200
300
400
500
600
With speed lanes
Without speed lanes
w
Ave
rage
tra
vers
al t
ime
Motion Planning for Multiple Autonomous Vehicles
Speed Lanes
rkala.99k.org
0100200300400500600700800900
1000
With dy-namic speed lanes
With fixed speed lanes
Speed Upper Bound
Ave
rage
tra
vers
al t
ime
Comparisons by increasing diversity – test the algorithm adaptability
Motion Planning for Multiple Autonomous Vehicles
Speed Lanes
rkala.99k.org
Comparisons by increasing diversity – test the algorithm adaptability
0
100
200
300
400
500
With dy-namic speed lanes
With fixed speed lanes
Speed Lower Bound
Ave
rage
tra
vers
al t
ime
Motion Planning for Multiple Autonomous Vehicles
Route Planning • How to choose between a shorter/congested
route or a longer/not congested route?
• Use a standard graph search on road network graph
• Re-plan at every crossing to cope with changing traffic
• Cost function: length + penalty x traffic density
• For near roads use Current traffic density• For further roads use predicted traffic densityrkala.99k.org
Motion Planning for Multiple Autonomous Vehicles
Rout Planning
rkala.99k.org
Entry Point
of vehicle
s
Exit Point
of vehicle
s
Straight Road
Longer Diversion
Shorter Diversion
Vehicles first use straight road, on congestion of which shorter diversion is
also used, on congestion of which, longer diversion is also used.
Motion Planning for Multiple Autonomous Vehicles
Route Planning
rkala.99k.org
240
260
280
300
320
340
360
380
Proposed systemDistance minimiza-tion
α
Ave
rage
tra
vers
al t
ime
Comparisons with distance minimization graph search (only straight road used)
by varying penalty
Motion Planning for Multiple Autonomous Vehicles
Route Planning• Lower penalty = Distance minimization graph
search• Lower penalty = Search which attempts to
equate density on all roads
For the considered map:• Lower penalty makes straight road congested,
poor performance• Higher penalty encourages vehicles to take
diversions even though main road may not be too congested, poor performance
rkala.99k.org
Motion Planning for Multiple Autonomous Vehicles
Booking• Reserve a road or lane for privileged set of
vehicles• Key issue: How many vehicles to be booked?
rkala.99k.org
Booking
Road Booking
A road/ section of road booked
E.g. VIP road for a concert
Lane Booking
One of the lanes
booked
E.g. Olympic Vehicles
Lane
Emergency vehicle may
book lanes as they go
Motion Planning for Multiple Autonomous Vehicles
Booking
rkala.99k.org
Booked road (only for privileged/booked
vehicles
Diversion for all other vehicles
Motion Planning for Multiple Autonomous Vehicles
Booking
Booking more vehicles makes it take longer for the booked vehicles and shorter for
the other vehiclesrkala.99k.org
0100200300400500600700800
Non-booked vehicles
Booked vehicles
No vehicle booked
Percent of vehicles booked
Ave
rage
tra
vers
al t
ime
Road Booking
Motion Planning for Multiple Autonomous Vehicles
Booking
Booking more vehicles makes it take longer for the booked vehicles and shorter for
the other vehicles rkala.99k.org
0
50
100
150
200
250
300
Non-booked vehicles
Booked ve-hicles
All vehicles booked
No vehicle booked
Percent booked vehicles
Ave
rage
tra
vers
al t
ime
Lane Booking
Motion Planning for Multiple Autonomous Vehicles rkala.99k.org
Thank You
• Acknowledgements:
• Commonwealth Scholarship Commission in the United Kingdom
• British Council