Receding Horizon Task and Motion Planning in Changing Environments

Complex manipulation tasks require careful integration of symbolic reasoning and motion planning. This problem, commonly referred to as Task Motion Planning (TAMP), is even more challenging if the workspace non-static, e.g. due human interventions perceived with noisy non-ideal sensors. work proposes an online approximated TAMP method that combines a geometric module planner standard task in receding horizon fashion. Our approach iteratively solves reduced planning problem over window limited number future actions during implementation actions. Thus, only first action actually scheduled at each iteration, then moved forward, solved again. procedure allows naturally take into account potential changes scene while ensuring good runtime performance. We validate our within extensive experiments simulated environment. showed able deal unexpected environment comparable performance respect other recent approaches solving traditional static benchmarks. release this paper open-source method.