Robot Task Planning on Multiple Environments

Abstract

This paper presents a hierarchical planning approach to solving the global localization problem for mobile robots using a compaction map method. This approach uses architectural design features such as walls and doors to break the environment into rooms and cluster them by similarity constraints. Each group of similar rooms is mixed into a single, compact representative map. Lighter POMDP plans are generated only for these compact maps and not for the whole environment, decreasing size of the set of possible states. In the higher layer of the planner, the robots use the compact maps to define which group of rooms they are located in. In the lower layer, the specific room of a representative map is isolated and its probability distribution is used to find out the precise localization. We demonstrate the hierarchical approach and the new map compaction with experiments on both V-REP Simulator and a Pioneer 3DX robot. This approach allows the robot to perform both the localization and tasking in a large-scale environment while keeping the accuracy.

Anticipative Shared Control for Robotic Wheelchairs Used by People with Disabilities

“…The destination can be setup by user and once the robot knows its pose (based on the probability distribution that has been updated by the POMDP model), a path planning based on the A* or D* will be used to reach the destination. Otherwise, the robot assumes to be in the higher probability state and chooses the best action toward the destination [12]. Even when the pose of the robot is well know, the localization planning keeps running constantly updating its probability distribution and consequently, its pose.…”