Constructing Discrete Event Abstractions for Planning and Controlling Mobile Robots

This paper presents a procedure for creating a probabilistic finite-state model for mobile robots and for finding a sequence of controllers ensuring the highest probability for reaching some desired regions. The approach starts by using results for controlling affine systems in simpliceal partitions, and then it creates a finite-state representation with history-based probabilities on transitions. This representation is embedded into a Petri Net model with probabilistic costs on transitions, and a highest probability path to reach a set of target regions is found. The proposed probabilistic framework is suitable for controlling mobile robots based on more complex specifications.