robot rendezvous: 3 or more robots using 1-dimensional search !!!!!!! !

43
ROBOT RENDEZVOUS: 3 OR MORE ROBOTS USING 1- DIMENSIONAL SEARCH ! ! ! ! ! ! ! !

Upload: casey-ellender

Post on 14-Dec-2015

225 views

Category:

Documents


2 download

TRANSCRIPT

  • Slide 1

ROBOT RENDEZVOUS: 3 OR MORE ROBOTS USING 1-DIMENSIONAL SEARCH !!!!!!! ! Slide 2 OVERVIEW Approaches Tested Search approaches ensuring coverage of the entire environment Final Approach Reduces problem to a 1-Dimensional Search Slide 3 APPROACHES TESTED Slide 4 GRID-BASED ITERATIVELY DEEPENING DEPTH FIRST SEARCH Slide 5 GRID-BASED IDDFS Slide 6 Based on "Symmetric Rendezvous in Planar Environments with and without Obstacles" - Isler, et. al. Decompose map into square cells of side length L Search each cell by counterclockwise perimeter circumnavigation Neighboring cells recursively searched on discovery up to a prescribed depth D. D increases each iteration by the square root of 2. Iterative deepening attempts to simulate optimality of breadth-first search. Slide 7 GRID-BASED IDDFS: SIMULATION VIDEO Slide 8 Slide 9 GRID-BASED IDDFS: CHALLENGES C-space obstacles not known beforehand. Need to use on-line wall following. Too large a cell size may cause areas to be exempt from the search For these reasons, we did not use this Corridor 1 (searched first) Corridor 3 (never searched) Corridor 2 (searched second) Slide 10 LOGARITHMICALLY EXPANDING SPIRAL SEARCH Slide 11 LOG SPIRAL SEARCH Based on "Spiral Search as an Efficient Mobile Robotic Search Technique" - Burlington, et. al. Generate a series of coordinates with increasing distance from origin for robot to goto When in contact with obstacles, wall follow using Bug2 Algorithm to navigate to next point Slide 12 Slide 13 Slide 14 Slide 15 Slide 16 Slide 17 Slide 18 Slide 19 Slide 20 Slide 21 LOG SPIRAL SEARCH: CHALLENGES Online wall following in an unknown environment has its challenges Areas may be completely bypassed if robot starts in a long tunnel Future work would require a different approach to logarithmic spiraling in order to bypass this problem. Slide 22 HIGHER-LEVEL RENDEZVOUS IN UNBOUNDED ENVIRONMENT LF1F3F2 Slide 23 HIGHER-LEVEL RENDEZVOUS IN UNBOUNDED ENVIRONMENT Slide 24 R1 R2 R3 R4 Slide 25 R1 R2 R3 R4 Slide 26 R1 R2 R3 R4 INTERMEDIATE RENDEZVOUS Slide 27 R1 R2 R3 R4 Slide 28 R1 R2 R3 R4 Slide 29 R1 R2 R3 R4 INTERMEDIATE RENDEZVOUS Slide 30 R1 R2 R3 R4 Slide 31 R1 R2 R3 R4 FINAL RENDEZVOUS Slide 32 R1 R2 R3 R4 Slide 33 1-DIMENSIONAL RENDEZVOUS IN BOUNDED ENVIRONMENT FINAL APPROACH METHODOLOGY LFFF Slide 34 METHODOLOGY: SYSTEM OVERVIEW Slide 35 METHODOLOGY: OUTER BOUNDARY USING DISTBUG Slide 36 METHODOLOGY: 1D SEARCH Slide 37 METHODOLOGY: HIGHER-LEVEL RENDEZVOUS R1 R2 R3 R4 Slide 38 METHODOLOGY: HIGHER-LEVEL RENDEZVOUS R1 RENDEZVOUS R3 R4 R2 COLLISION AVOIDANCE Slide 39 METHODOLOGY: SIMULATION VIDEO Slide 40 METHODOLOGY: COMPETITIVE RATIO Slide 41 METHODOLOGY: ADVANTAGES Its a very elegant and compact algorithm Search area is greatly reduced (so we don't need to guarantee coverage of the environment) Localization isn't as crucial after the outer boundary has been found (thus much easier to apply to real world applications) Less parameter adjustment is needed to apply to real world applications Slide 42 METHODOLOGY: CONSTRAINTS Method works only for bounded environment There is an upper bound on the number of robots that can perform the search with adequate "personal space" (deadlock) Rendezvous area is n*R (where R is the communication range and n is the number of robots) We can't guarantee coverage of the environment Slide 43 FURTHER WORK Add communication between colliding robots to lessen the time consumed by 1D search Better search for outer bound Decrease rendezvous area Test algorithm with more robots