`

Timezone: »

 
Poster
Successor Feature Landmarks for Long-Horizon Goal-Conditioned Reinforcement Learning
Christopher Hoang · Sungryull Sohn · Jongwook Choi · Wilka Carvalho · Honglak Lee

Thu Dec 09 04:30 PM -- 06:00 PM (PST) @ Virtual

Operating in the real-world often requires agents to learn about a complex environment and apply this understanding to achieve a breadth of goals. This problem, known as goal-conditioned reinforcement learning (GCRL), becomes especially challenging for long-horizon goals. Current methods have tackled this problem by augmenting goal-conditioned policies with graph-based planning algorithms. However, they struggle to scale to large, high-dimensional state spaces and assume access to exploration mechanisms for efficiently collecting training data. In this work, we introduce Successor Feature Landmarks (SFL), a framework for exploring large, high-dimensional environments so as to obtain a policy that is proficient for any goal. SFL leverages the ability of successor features (SF) to capture transition dynamics, using it to drive exploration by estimating state-novelty and to enable high-level planning by abstracting the state-space as a non-parametric landmark-based graph. We further exploit SF to directly compute a goal-conditioned policy for inter-landmark traversal, which we use to execute plans to "frontier" landmarks at the edge of the explored state space. We show in our experiments on MiniGrid and ViZDoom that SFL enables efficient exploration of large, high-dimensional state spaces and outperforms state-of-the-art baselines on long-horizon GCRL tasks.

Author Information

Christopher Hoang (University of Michigan)
Sungryull Sohn (LG AI research US)
Jongwook Choi (University of Michigan)
Wilka Carvalho (University of Michigan)

I am a Masters student and NSF Graduate Research Fellow in the Computer Science Department at the University of Southern California. My primary research interest is the development of neuroscience- and cognitive science-informed artificial intelligence and machine learning models with brain-rivaling information-processing capabilities.

Honglak Lee (U. Michigan)

More from the Same Authors