Nick Hawes Seminar

For our autonomous robots to be useful they must act to complete missions, i.e. sequences of actions created to achieve goals or service tasks. However, robotic action execution is not deterministic, therefore techniques for mission planning  under uncertainty are required for autonomous systems. In this talk I will motivate the combination of Markov decision processes and probabilistic verification as a framework for robot mission planning under uncertainty. I will motivate this using examples of mobile service robots capable of long-term autonomy in everyday environments. Following this I will present some of our recent work on extending this framework to embed reinforcement learning of actions, human interventions, multiple mission objectives, and multiple robots.

Mission planning for long-horizon tasks requires the planning agent to use a model to encode its interaction with its environment. In most robotic tasks some parts of this model are known with certainty, whereas other parts may only be known with uncertainty at design time, and must be updated via learning either between missions (i.e. “offline") or during execution (“online"). In this talk I’ll give a high-level summary of our recent work on mission planning with such uncertain models. This will range from planning in MDPs with a Gaussian Process prior over a single state features,  to planning in Uncertain MDPs and Bayes-Adaptive MDPs where the true model cannot be known with certainty.