Efficient Symbolic Task Planning for Multiple Mobile Robots
Yuqian Jiang · 2016
Abstract : Symbolic task planning enables a robot to make high-level deci-sions toward a complex goal by computing a sequence of actions withminimum expected costs. This thesis builds on a single-robot planningframework, and aims to address two issues: (1) lack of performance in-formation in the selection of planners across dierent formalisms, and(2) time complexity of optimal planning for multiple mobile robots. Inthis thesis we rst investigate the performance of the state-of-the-artsolvers of Planning Domain Denition Language (PDDL) and AnswerSet Programming (ASP) in robot navigation problems. We then aim toreduce overall costs where multiple mobile robots may block in narrowcorridors or collaborate to open doors. It is challenging to model suchinteractions due to uncertain delays of navigation actions in populatedareas. This paper addresses this challenge with an algorithm which cal-culates the conditional distribution of plan costs for each robot, givenplanned actions of other robots. We then propose an iterative con-ditional planning algorithm to eciently approximate optimal plansof the system. Experiments in simulation and a demonstration onreal robots show that the algorithm has a signicant advantage overbaselines in which robots plan individually, or plan together without amodel for navigation action durations.