As more autonomous agents share space with people, it is crucial that these agents are cognizant of how their behaviors are interpreted. Previous works have investigated various methods to communicate or conceal agents' goals, intentions, and capabilities through their behaviors. This thesis presents a unifying model for generating behaviors that not only achieve desired goals but also consider how these behaviors are interpreted. The proposed model, known as the OAMDP, operates under the assumption that observers interpret the agent's actions to form beliefs about the agent's potential desires, goals, and intentions. Planning with OAMDP then yields behaviors that lead to desirable beliefs in the observer. While OAMDP is an expressive framework capable of producing various kinds of observer-aware behaviors, reasoning about the observer's beliefs introduces a dependence on histories, which makes solving OAMDPs intractable in the worst case. To address this issue, the thesis investigates several approximation algorithms for OAMDPs. Finally, it discusses future directions, including a constraint-based formulation of the problem and the use of deeper levels of nested reasoning.
Advisor: Shlomo Zilberstein