In this project we consider a path-planning framework for first rescuers in an emergency building evacuation. The task for the rescuer is to reach multiple predefined sensitive areas of the building and then leave the building while avoiding a spreading hazard within a finite time horizon. We design a recursive path-planning algorithm with the objective of maximizing the probability of fulfilling the navigation task. The rescuers and the hazard are modeled as stochastic processes, whereas the sets to be reached are static and its location is assumed to be known. The algorithm is based on automata satisfaction formulated as a stochastic reachability problem with random sets. By using the modes of an automata, we formulate the problem as an automaton specification satisfaction problem. We augment the model of the rescuer with discrete modes, which capture information on the progress of the task accomplishment. We apply dynamic programming to obtain the optimal navigation policy and its corresponding probability value of fulfilling the task. This is tested in simulation and different extensions are applied later to make the process more realistic.

Outgoing