Closing The Loop Of Task Planning, Action And Sensing

Abstract

A bstract - This paper deals with the implementation of an Intelligent Assembly Robotic System incorporating an interface with Computer Vision, an efficient Task Planning methodology as well as task execution and monitoring through sensing. Two possibilities are available for task plan generation: the generation of the fastest plan and the generation of the best plan. The great efficiency of the method is mainly due to the choice of good heuristics and the early pruning of unnecessary paths in the plan generation tree. This system analyses and solves some non-trivial problems including g eometric c onstraints, i ntelligent object h andling, s pecific o perations (e.g. r egrasp ing) , simplified collision avoidance and management of the working area. Low level sensing has been included for task monitoring. Their use also provides reactive capabilities to the system. Recovery from unexpected situations can be achieved.