Information-based Exploration Strategy for Mobile Robot in Dynamic Environment

To meet the necessity of handling environmental uncertainties of mobile robots, we proposed an efficient exploration strategy to gather information, called Entropy Sweeper. To do so, we utilized the entropy distribution and the utility function to determine which positions have more uncertainties. Proposed strategy is divided into two phases: the learning phase and the action phase. In general, uncertainties increase unevenly and never disappear in dynamic environments. So in the learning phase, robots move wall to wall to learn which positions are likely to increase uncertainties actively. In the action phase, robots explore the environment efficiently and continue lifelong learning to handle environmental uncertainties. This strategy is an optimization not only for paths but also for sequences of exploration points using information about uncertainties of dynamic environments. We demonstrated its effectiveness with several simulations.