Timezone: »

 
Poster
Policy Optimization with Linear Temporal Logic Constraints
Cameron Voloshin · Hoang Le · Swarat Chaudhuri · Yisong Yue

Tue Nov 29 09:00 AM -- 11:00 AM (PST) @ Hall J #719
in Poster Session 1 »

We study the problem of policy optimization (PO) with linear temporal logic (LTL) constraints. The language of LTL allows flexible description of tasks that may be unnatural to encode as a scalar cost function. We consider LTL-constrained PO as a systematic framework, decoupling task specification from policy selection, and an alternative to the standard of cost shaping. With access to a generative model, we develop a model-based approach that enjoys a sample complexity analysis for guaranteeing both task satisfaction and cost optimality (through a reduction to a reachability problem). Empirically, our algorithm can achieve strong performance even in low sample regimes.

Keywords: Optimization · Reinforcement Learning · linear temporal logic · RL · learning · Constrained · LTL · Policy

Author Information

Cameron Voloshin (California Institute of Technology)
Hoang Le (Argo AI)
Swarat Chaudhuri (The University of Texas at Austin)
Yisong Yue (Caltech)

More from the Same Authors