Online Robot Task Switching Under Diminishing Returns

We investigate the task switching problem of a robot maximizing its long-term average rate of return on work performed. We propose an online method to maximize the average gain rate based on only past experience. For that we alter the formulation from optimal foraging theory and recursively include estimates of global task qualities. We demonstrate and analyze our method on a puck-foraging example. In simulation experiments under a variety of conditions we show that our method performs well compared to results obtained by brute force method using post-processed foraging data.