Skip to yearly menu bar Skip to main content


Poster

Composing Complex Skills by Learning Transition Policies

Youngwoon Lee · Shao-Hua Sun · Sriram Somasundaram · Edward S Hu · Joseph Lim

Great Hall BC #69

Keywords: [ reinforcement learning ] [ continuous control ] [ hierarchical reinforcement learning ] [ modular framework ]


Abstract:

Humans acquire complex skills by exploiting previously learned skills and making transitions between them. To empower machines with this ability, we propose a method that can learn transition policies which effectively connect primitive skills to perform sequential tasks without handcrafted rewards. To efficiently train our transition policies, we introduce proximity predictors which induce rewards gauging proximity to suitable initial states for the next skill. The proposed method is evaluated on a set of complex continuous control tasks in bipedal locomotion and robotic arm manipulation which traditional policy gradient methods struggle at. We demonstrate that transition policies enable us to effectively compose complex skills with existing primitive skills. The proposed induced rewards computed using the proximity predictor further improve training efficiency by providing more dense information than the sparse rewards from the environments. We make our environments, primitive skills, and code public for further research at https://youngwoon.github.io/transition .

Live content is unavailable. Log in and register to view live content