CRAMm -- Memories for Robots Performing Everyday Manipulation Activities

摘要: Agents that learn from experience can profit immensely memorizing what they have done, why, how, and happened. For autonomous robots performing complex manipulation tasks, these memories include low level data, such as perceptual snapshots of relevant scenes influenced decision making, detailed motions the robot performed, effects motions. They also high representations intended actions belief-dependent descisions led to chosen course action. In this paper, we propose CRAMm, a memory management system record very comprehensive informative without slowing down operation robot. CRAMm offers query interface allows retrieve kinds information stated above. This is done using first-order logical language provides predicates concerning beliefs intentions robot, its physical state, information, action their relations at various different levels abstraction.