SLOPE: Optimistic Potential Landscape Shaping for Model-based Reinforcement Learning

TL;DR

SLOPE introduces a novel reward shaping framework for model-based reinforcement learning that constructs informative potential landscapes to improve exploration and performance in sparse reward settings.

Contribution

It proposes a new method using optimistic distributional regression to create potential landscapes, enhancing exploration in sparse reward environments.

Findings

SLOPE outperforms baselines across 30+ tasks and 5 benchmarks.

It effectively handles fully sparse, semi-sparse, and dense reward scenarios.

Demonstrates success in real-world robotic deployments.

Abstract

Model-based reinforcement learning (MBRL) is sample-efficient but struggles in sparse reward settings. A critical bottleneck arises from the lack of informative gradients in sparse settings, where standard reward models often yield flat landscapes that struggle to guide planning. To address this challenge, we propose Shaping Landscapes with Optimistic Potential Estimates (SLOPE), a novel framework that shifts reward modeling from predicting sparse scalars to constructing informative potential landscapes. SLOPE employs optimistic distributional regression to estimate high-confidence upper bounds, which amplifies rare success signals and ensures sufficient exploration gradients. Evaluations on 30+ tasks across 5 benchmarks and real-world robotic deployments, demonstrate that SLOPE consistently outperforms leading baselines in fully sparse, semi-sparse, and dense rewards.