Download presentation
Presentation is loading. Please wait.
1
Providing Haptic ‘Hints’ to Automatic Motion Planners Providing Haptic ‘Hints’ to Automatic Motion Planners by Burchan Bayazit Department of Computer Science Texas A&M University
2
Motion Planning Why a computer? –It is fast –It is patient and does not complain Why a human? – He/she can really(!) understand the problem Why a computer? –It is fast –It is patient and does not complain Why a human? – He/she can really(!) understand the problem GIVEN: an initial position and orientation and a goal position and orientation of an object A, FIND: A path T with continuous sequence of positions and orientations of A avoiding contact with other objects GIVEN: an initial position and orientation and a goal position and orientation of an object A, FIND: A path T with continuous sequence of positions and orientations of A avoiding contact with other objects
3
Outline Configuration Space and probabilistic motion planners Haptic Environments Human/Computer Cooperation Conclusion Configuration Space and probabilistic motion planners Haptic Environments Human/Computer Cooperation Conclusion
4
Configuration Space Each point in configuration space corresponds the robot’s position and orientation in the workspace Robot
5
Workspace vs C-Space Workspace C-Space
6
Obstacles
7
C-Space Obstacle
8
Initial Goal
9
Roadmap
10
Automatic Motion Planners They are good, but sometimes fail in difficult environments How humans can contribute? –Give hints –Select important points –Connect some parts of roadmap by hand –Show an approximate path (not necessarily collusion free) They are good, but sometimes fail in difficult environments How humans can contribute? –Give hints –Select important points –Connect some parts of roadmap by hand –Show an approximate path (not necessarily collusion free)
11
Haptic Interface User can feel the environment User can move the robot around the obstacles, by actually feeling when robot touches something User can select a specific configuration of robot User can feel the environment User can move the robot around the obstacles, by actually feeling when robot touches something User can select a specific configuration of robot relating to or based on the sense of touch
12
Phantom
13
Proposed System
14
Applying Force
15
Penetration
16
Ideal Algorithm Robot Position Haptic Loop Calculate Force Return Force Uses penetration distance Must have 1Khz update rate Uses penetration distance Must have 1Khz update rate
17
PROBLEM: Penetration is not available for complex environments SOLUTION: If the robot is in collision, push outside with a constant force PROBLEM: Penetration is not available for complex environments SOLUTION: If the robot is in collision, push outside with a constant force
18
Collision Detection PROBLEM: It is too slow (~50Hz on SGI O2) SOLUTION: Use a distributed environment, i.e. use a faster computer for computations Use heuristic to determine when collision occurs PROBLEM: It is too slow (~50Hz on SGI O2) SOLUTION: Use a distributed environment, i.e. use a faster computer for computations Use heuristic to determine when collision occurs
19
Heuristically decide collision Robot Position Get the result Calculate Collision Haptic Loop Computation Loop Yes No Previous result ready? Collision ? Yes No Return No ForceReturn Constant Force Request a new calculation
20
P P LF Last free cfg (computed) Minimum distance (md) P P LF md Current distance (cd) if projection of cd > md then collision else free
21
P P LF LC Last colliding cfg (computed) P P P LF LC P collision distance (xd) Current distance (cd) if projection of cd > xd then collision else free
22
Roadmap Visualization and Sensing Displaying configurations Displaying paths/roadmaps –Display roadmap edges –Display individual roadmap configurations Feeling paths/roadmaps Displaying configurations Displaying paths/roadmaps –Display roadmap edges –Display individual roadmap configurations Feeling paths/roadmaps
23
Environment
24
Roadmap Edges
25
Edges and Cfgs
26
Human/Planner Cooperation Human selects critical configurations Human selects a path Computer improves its roadmap using these configurations Approximate path (easier for the human)
28
Approximate Path Approximate path generated by human Approximate path generated by human C-Space Obstacle
29
Push Towards Line Segment C-Space Obstacle
30
Push Toward the Closest Surface cfg C-Space Obstacle
31
Push in Workspace (using vertices) Workspace Obstacle Robot d4 d1 d3 d2 Pair each vertex of robot with workspace obstacle Move in the closest direction Pair each vertex of robot with workspace obstacle Move in the closest direction
32
Push in Workspace (using normals) Workspace Obstacle Robot Find Colliding Segments on robot and the obstacle Translate robot so that those segments face each other. Find Colliding Segments on robot and the obstacle Translate robot so that those segments face each other.
35
Application Areas Any motion planning problem Any master-slave problem Checking engine parts separability in the design phase Robotics surgery with the supervision of a surgeon
36
Conclusion We described automated motion planners We described a haptic interface for automated motion planners –Heuristic for force feedback in crowded areas We showed how user can cooperate with the planner –Roadmap Visualization –Approximate path We described automated motion planners We described a haptic interface for automated motion planners –Heuristic for force feedback in crowded areas We showed how user can cooperate with the planner –Roadmap Visualization –Approximate path
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.