A multi - robot system for adaptive exploration ofa fast changing environment : probabilisticmodeling and experimental

Is it more e cient to use one or several robots? Will the performance of a group of robots working in a collaborative task be enhanced if the robots can communicate with one another? What learning abilities should the robot(s) be provided with for adapting to a continuously changing environment? We address these three issues in a speci c task, namely learning the topography of an environment whose features change frequently. We propose a theoretical framework based on probabilistic modeling to describe the system's dynamics. The adaptive multirobot system and its dynamic environment are modeled though a set of probabilistic equations. The model gives an explicit description of the in uence of the variables of the system, namely the number of worker robots, the frequency of environmental changes and the environment's con guration, on the data collecting performance of the group. It is then used to determine boundaries for these system's variables within which the learning is successful. Further, we implement the multi-robot system in experiments with a group of Khepera robots and in simulation using Webots, a 3-D simulator of the Khepera robots. The robots are controlled by a distributed architecture. Each robot's controller is based on an associative memory type of learning algorithm. Results show that the algorithm allows a group of robots to keep an up-to-date account of the environmental state when this changes regularly. Finally, the predictions of the probabilistic model are compared to the results of simulations and physical experiments. It is found that the model shows both a good qualitative and quantitative correspondence to these results. This suggests that a probabilistic model can be a good rst approximation of a multi-robot system when this system's behavior is stochastic in nature.