220 likes | 341 Views
Autonomous construction of maps by miniature robots. Paul Fitzpatrick Colin Flanagan. Robot’s path traces the outline of the boundary. To follow a boundary :- Move forward continuously Turn left if boundary gets further away Turn right if boundary gets closer. Corner feature. Move left.
E N D
Autonomous construction of mapsby miniature robots Paul Fitzpatrick Colin Flanagan
Robot’s path traces the outline of the boundary • To follow a boundary :- • Move forward continuously • Turn left if boundary gets further away • Turn right if boundary gets closer Corner feature Move left Move right Edge feature
Corner Landmarks Landmark position Landmark angle Landmark position • Location of corners used to correct robot’s position estimate • Angle between pairs of corners allow corrections to direction estimate
Concave corner Convex corner Robot turns more sharply than the curve itself, giving a well-defined landmark. Robot cannot turn an arc with a radius less than the robot itself.
Second pass First pass Wall dq Straight-Edge Landmarks dq • Correction suggested by apparent edge position • Correction suggested by apparent direction of the edge
“Marker” representation scheme Recent marker Path of robot
Recent marker Older marker Replaced marker Path of Robot Updating markers
Recent marker Older marker Replaced marker Path of Robot Map = collection of markers
Neighbourhoods Immediate Neighbourhood Near Neighbourhood Local Neighbourhood Distant Neighbourhood Current position of robot Markers
Immediate Near Local Distant Index to marker currently being sorted Index of next marker to be sorted In each cycle of the sorting process, one marker from each neighbourhood is examined and moved to another neighbourhood if appropriate.
Cyclic behaviour Robot caught in cyclic behaviour Robot shown trapped by an awkwardly shaped obstacle.
Use of familiarity Familiar region expands until escape is possible Robot uses “familiarity” sensor to escape from trap.
Use of background search Inspiration- background search succeeds Use of familiarity can be combined with conventional search
Connectivity Markers not laid successively cannot be assumed to be reachable from each other, even if they are close Markers placed successively as robot moves can be assumed reachable from each other A B Path of robot
Deducing connectivity Markers close to each other can be deduced to be connected if a short path between them can be found through other connected markers Path of robot A B C D D
Direction of spread Most reachable Least reachable Updating connectivity
Using connectivity 3 4 3 3 2 3 1 2 1 Goal 1 2 2 2 3 3
Autonomous construction of mapsby miniature robots Paul Fitzpatrick Colin Flanagan
Conclusions • Maps can be built by an autonomous robot even under very constrained conditions: • Without long-range sensors such as sonar or vision • With limited processing power • With limited memory • Maps can be suitable for use by “behaviour-based” robots • No shared representation necessary • Don’t require extensive, time-extended computation
Second pass First pass Wall dq Straight-Edge Landmarks dq • Correction suggested by apparent edge position • Correction suggested by apparent direction of the edge