Mixing differential inclusions with Markov decision processes

This paper describes a behavioral approach to the synthesis of controllers for the execution of tasks by mobile robots. The paradigm integrates two distinct methods for decision making under uncertainty, namely Markov decision processes and differential inclusions. The target application is surveillance systems where mobile robots can approach intruders to take their mugshot or block exit pathways. An hybrid model of the system is described and stability conditions discussed. The main ideas are exemplified in a search and pursue application and some preliminary experimental results with a navigation behavior are also presented.