RISE: Self-Improving Robot Policy with Compositional World Model

TL;DR

RISE introduces a self-improving robotic reinforcement learning framework using a compositional world model that predicts future states and evaluates imagined outcomes, enabling efficient policy improvement without physical environment resets.

Contribution

The paper presents a scalable, imagination-based RL framework with a compositional world model that enhances robot policy learning in contact-rich tasks.

Findings

Achieved over +35% performance in dynamic brick sorting.

Improved backpack packing success rate by +45%.

Enhanced box closing performance by +35%.

Abstract

Despite the sustained scaling on model capacity and data acquisition, Vision-Language-Action (VLA) models remain brittle in contact-rich and dynamic manipulation tasks, where minor execution deviations can compound into failures. While reinforcement learning (RL) offers a principled path to robustness, on-policy RL in the physical world is constrained by safety risk, hardware cost, and environment reset. To bridge this gap, we present RISE, a scalable framework of robotic reinforcement learning via imagination. At its core is a Compositional World Model that (i) predicts multi-view future via a controllable dynamics model, and (ii) evaluates imagined outcomes with a progress value model, producing informative advantages for the policy improvement. Such compositional design allows state and value to be tailored by best-suited yet distinct architectures and objectives. These components are integrated into a closed-loop self-improving pipeline that continuously generates imaginary rollouts, estimates advantages, and updates the policy in imaginary space without costly physical interaction. Across three challenging real-world tasks, RISE yields significant improvement over prior art, with more than +35% absolute performance increase in dynamic brick sorting, +45% for backpack packing, and +35% for box closing, respectively.