Timezone: »

 
Poster
Learning Task Specifications from Demonstrations
Marcell Vazquez-Chanlatte · Susmit Jha · Ashish Tiwari · Mark Ho · Sanjit Seshia

Wed Dec 05 02:00 PM -- 04:00 PM (PST) @ Room 517 AB #166
in Wed Poster Session B »

Real-world applications often naturally decompose into several sub-tasks. In many settings (e.g., robotics) demonstrations provide a natural way to specify the sub-tasks. However, most methods for learning from demonstrations either do not provide guarantees that the artifacts learned for the sub-tasks can be safely recombined or limit the types of composition available. Motivated by this deficit, we consider the problem of inferring Boolean non-Markovian rewards (also known as logical trace properties or specifications) from demonstrations provided by an agent operating in an uncertain, stochastic environment. Crucially, specifications admit well-defined composition rules that are typically easy to interpret. In this paper, we formulate the specification inference task as a maximum a posteriori (MAP) probability inference problem, apply the principle of maximum entropy to derive an analytic demonstration likelihood model and give an efficient approach to search for the most likely specification in a large candidate pool of specifications. In our experiments, we demonstrate how learning specifications can help avoid common problems that often arise due to ad-hoc reward composition.

Keywords: Relational Learning · Markov Decision Processes

Author Information

Marcell Vazquez-Chanlatte (University of California, Berkeley)
Susmit Jha (SRI International)
Ashish Tiwari (SRI International)
Mark Ho (UC Berkeley)
Sanjit Seshia (UC Berkeley)

More from the Same Authors