Inferring helpful actions via inverse reinforcement learning and makespan optimization
Recent developments in household robotics, and the emergence of new personal robotic platforms show encouraging progress toward automating less desired household tasks, like organization, cleaning, kitchen preparation, and household maintenance. Effective communication with robots will help reduce the burden required to manage these tasks. Natural language commands as a communication methodology are often ambiguous and insufficiently clear, and an understanding of the human's intentions are required to disambiguate correctly. This project seeks to develop a method to allow a robot to reason about a human's intentions and choose actions which reduce the amount of time spent completing a collaborative task.