Difficulty-Aware Planning for Time-Bounded Mobile Robot Missions
We consider planning problems where a robot must gather reward by completing tasks at each of a large set of locations while constrained by a time bound. Our focus is problems where the difficulty of each task, and thus its duration, can be predicted, but is not fully known in advance. We model difficulty-aware problems as a Markov decision process with discrete time in the state, and propose variants on this model which allow adaptation to different robotics domains. Due to the intractability of the general model for difficulty-aware problems, we propose simplifications to allow planning in large domains. The key idea behind these simplifications is constraining navigation using a solution to the travelling salesperson problem. We evaluate our models on maps generated from real-world environments and consider two domains with different characteristics: UV disinfection, and cleaning. We evaluate the effect of model variants and simplifications on performance, and show that policies obtained for our models outperform a rule-based baseline, as well as a model which does not consider difficulty. We also evaluate our models in a real robot experiment where a quadruped performs simulated inspection tasks in an industrial environment.