Conclusions
Rahul Kala
April, 2013 School of Systems, Engineering, University of Reading rkala.99k.org
Conclusions
Thesis
Trajectory Generation
rkala.99k.org
2. 3. 4.
Locally optimal, Globally optimal for Near-Complete simple cases Generally optimal. Can miss overtakes with very fine turns Near-Complete
5. 6. 7. 8. 9.
Planning using Generally optimal. Can miss dynamic distributed overtakes with very fine turns lanes
Fuzzy Logic No
Near-Complete No
Very Low, Completely Scalable, N/A Very Low, Completely Scalable, Lateral Potentials No No N/A Generally optimal. Can miss very fine Near-Complete (less Medium, Very scalable (more Elastic Strip turns than 2, 3, 4 and 5) than 2 and 3), N/A Low, Almost completely Logic based planning Locally near-optimal. (less than 3) No (more than 6 and 7) scalable, N/A
rkala.99k.org
Computational time (least to highest): Fuzzy Logic, Lateral Potentials, Logic Based Planning, Elastic Strip, Multi Level Planning, Planning using Dynamic Distributed Lanes, RRTConnect, RRT and GA.
Motion Planning for Multiple Autonomous Vehicles rkala.99k.org
S. No.
Communication, Assumptions
Optimality
Computational complexity
1.
Somewhat high to Sub-optimal, Yes, Vehicles stay continuously alter speed and Traffic inspired heuristics for Global knowledge on their left check overtake feasibility. path/speed, Prioritization makes it more sides mostly Computation is distributed as desirable the vehicle travels Non-cooperative, Sub-optimal Non-cooperative, Sub-optimal A little high due to multiple attempts to compute speed Small time needed to decide between overtaking and vehicle following
2.
Prioritization, Attempt to Rapidly-Exploring Yes, One way maintain maximum collisionRandom Trees (RRT) traffic only free speed RRT-Connect Prioritization, Vehicle following/ overtake based speed determination Yes, One way traffic only
3.
4.
Layered Prioritization, Each layer uses separation Multi Level Planning maximization heuristic, Yes Vehicle following/ overtaking based speed determination Pseudo-centralized, Each state expansion uses Planning using separation maximization dynamic distributed Yes heuristic, Vehicle following/ lanes overtaking based speed determination Motion Planning for Multiple Autonomous Vehicles
Largely optimal
5.
rkala.99k.org
Communication, Assumptions No, Vehicles stay on their left sides mostly, Roads not too wide to accommodate multiple vehicles per side of travel
Optimality
Computational complexity
7.
Lateral Potentials
8.
Elastic Strip
Vehicles treated as obstacles, Always overtake strategy, Distance from No, One way only front used for deciding speed Vehicles treated as moving obstacles, Always overtake strategy, Distance from No, One way only front used for deciding speed Vehicles treated as moving obstacles, Lateral distances No, Vehicles stay on measured for overtake their left sides decision making, Distance mostly from front used for deciding speed
Sub-optimal, Not accounting for Nil global knowledge makes it undesirable Sub-optimal, Not accounting for Very small time needed to global knowledge extrapolate vehicle motion makes it undesirable Sub-optimal, Cooperation can be slow, Not Very small time needed to accounting for extrapolate vehicle motion global knowledge makes it undesirable
9.
rkala.99k.org
Computational Expense
Cooperation
Overtaking
Speed Determination
Deliberative
Cooperative
Always overtake
Immediate best
Reactive
Noncooperative
Compute feasibility
Optimized assignment
rkala.99k.org
2. 3. 4.
cyclic, earliest vehicle first based, most late vehicles first based overtake based (extra lane primarily used for overtaking), cooperative to vehicles running more 5. Lane change late, dynamic speed limit based, booked lane (travel cost) 6. Traffic for Multiple Autonomous entirely semi-autonomous, mixed, manual rkala.99k.org Motion Planning Vehicles
Thank You
Motion Planning for Multiple Autonomous Vehicles rkala.99k.org