Clearance Based Path Optimization for Motion Planning Roland Geraerts and Mark Overmars ICRA 2004
Dec 16, 2015
Problem setting Low quality paths
Often long detours Discontinuity Unnecessary motions
Quality criteria Length Clearance Combination
Path Quality – Length
Path pruning Fast and simple heuristic Translational dist. improves
considerably Rotation is more difficult to remove
Path Quality – Length
Weakness of the shortcut heuristic Path can still contain many redundant
(rotational) motions Creating shortcuts on large portions
will fail Partial shortcut
Query path Shortcut Partial shortcut
Path Quality – Clearance Method
Remove redundant nodes Retract path to medial axis Remove branches
Can help improving path length
Query path Retracted path No branches
Path Quality – Combination Technique
Retract path to medial axis Increase size of robots Create partial shortcuts
Result Reasonable short path Path has a particular minimum
amount of clearance where it is possible
Experimental Setup – Path Length
Environment SAMPLE 3 test scenes/paths
Focus Free flying objects Post processing
Experiments – Simple Corridor Simple scene, cylinder Many motions are redundant Redundant motions are easy to remove
Experiments – Simple Corridor Simple scene, cylinder Many motions are redundant Redundant motions are easy to remove
0
200
400
600
Initial path Redundantnodes
Shortcut MA +Shortcut
Partialshortcut
MA+Partialshortcut
Optimal
Pat
h l
eng
th
translation
rotation400%
27% 16% 14% 1% 2% 0%
Experiments – Corridor Elbow shaped object is forced to rotate Little clearance to corridors
0
250
500
750
Initial path Redundantnodes
Shortcut MA +Shortcut
Partialshortcut
MA+Partialshortcut
Optimal
Pat
h l
eng
th
translation
rotation
256%
65%24% 24
%11% 5% 0%
Experiments – Hole Object must rotate to get through the
hole Clearance is small inside the hole
0
40
80
120
Initial path Redundantnodes
Shortcut MA + Shortcut Partial shortcut MA+Partialshortcut
Optimal
Pat
h le
ng
th
translation
rotation
41% 14% 0%
184%
112%
36%
23%