@inproceedings{MJlta18,
    author = {Abdel-Illah Mouaddib and Laurent Jeanpierre},
    title = {A Practical Distributed Knowledge-Based Reasoning and Decision-Theoretic Planning for Multi-robot Service Systems},
    booktitle = {Proceedings of the AAAI Fall Symposium on Reasoning and Learning in Real-World Systems for Long-Term Autonomy},
    pages = {55--61},
    year = {2018},
    abstract = {This paper presents a practical model of distributed reasoning and planning for a fleet of robots serving people in a shopping mall using distributed knowledge-based reasoning and distributed Markov Decision Process (MDP) where the environment changes frequently and the set of goals is not static. The model we present, in this paper, consists of distributed local reasoning and planning where each robot locally reasons on its perceived data (locally: onboard cameras and also from global perception system: external cameras) to update its local Knowledge Base (KB). Local KBs derive local goals and the local planners select the goal to accomplish and compute the policy to accomplish it while maintaining a coordinated behavior with the other robots by avoiding conflicts on goals. To this end, we propose a distributed market-based auction planning algorithm using a regret and opportunity costs in a distributed value function leading to augmented MDPs to coordinate the robots and to select the appropriate goals to accomplish. Our approach assumes communication between robots and external sensor and we will describe a method to minimize the dis-coordination (conflits on goals) when the communication is lacking. Experimental results on the algorithm performance and the implementation on real service robots in a shopping mall showed a very satisfying behavior as shown in the video.}
}