Implicit Robot-Human Communication in Adversarial and Collaborative Environments
Users of AI systems may rely upon them to produce plans for achieving desired objectives. Such AI systems should be able to compute obfuscated plans whose execution in adversarial situations protects privacy as well as legible plans which are easy for team-members to understand in collaborative situations. We develop a unified framework that addresses these dual problems by computing plans with a desired level of comprehensibility from the point of view of a partially informed observer. Our approach produces obfuscated plans with observations that are consistent with at least 'k' goals from a given set of decoy goals. In addition, when the goal is known to the observer, our approach generates obfuscated plans with observations that are diverse with at least 'l' candidate plans. Our approach for plan legibility produces plans that achieve a goal while being consistent with at most 'j' goals in a given set of confounding goals. We provide an empirical evaluation to show the feasibility and usefulness of our approaches.
READ FULL TEXT