LTL2Action: Generalizing LTL Instructions for Multi-Task RL
Pashootan Vaezipoor * 1 2 Andrew C. Li * 1 2 Rodrigo Toro Icarte 1 2 Sheila McIlraith 1 2 3
Abstract approaches do not scale well because they require (for every
We address the problem of teaching a deep rein- possible environment) manually building a large training
forcement learning (RL) agent to follow instruc- set comprised of natural language instructions with their
tions in ...


雷达卡




京公网安备 11010802022788号







