Multi-robot task allocation for collaborative adaptive informative sampling in structured environments

We look to improve the efficiency of teams of autonomous robots for information gathering approaches. These approaches are useful for environmental modeling, persistent monitoring, or mapping chemical signals in the air. An effective way of creating spatial models is by using Gaussian Process (GP) regression. Information-theoretic metrics, such as entropy or mutual information, can be used on top of the GP model’s predictions to decide where to place sensors or robots. In such case we speak of informative sampling. When the robot is creating such a model while sampling data this is called on-line or adaptive informative sampling. In this work we present our late breaking results extending our prior work [1] on multi-robot adaptive information gathering in structured indoor environments. This previous work implemented an information gathering rate adaptive algorithm to task robots with sensing an environment and building a GP model of RSSI signal readings. Multiple robots were coordinated by partitioning the environment into a number of segments equal to the number of robots and assigning each robot to its own region. In this paper we present a method for oversegmenting the sampling space into more regions than robots. A task assignment algorithm allows a central controller to dynamically task robots to different partition regions.