Dealing With Geometric Uncertainty Constraints When Planning Robot Actions

. A robot and its working space constitute a complex mechanical system that cannot be completely modelled. This means that the effects of robot actions cannot be exactly known at planning time, and that the resulting uncertainty may cause the generated robot control program to fail. Avoiding such failures requires determining the points where uncertainty is too high with regard to the required precision, and then to reduce this uncertainty using appropriate sensory based strategies. The purpose of this paper is to analyse the related problems and to describe the way we solved them. We will first present a method to compute and propagate the position uncertainty terms in a manipulation plan in order to determine the possible failure points. Then, we will focus on a specialized planner which allows the system to automatically generate the fine motion strategies which are required to amend the nominal manipulation plan. 1 Introduction A robot and its working space constitute a complex me..