EnerVerse-AC: Envisioning Embodied Environments with Action Condition

TL;DR

EnerVerse-AC introduces an action-conditional world model that generates realistic future visual observations for robotic tasks, reducing reliance on physical robots and complex simulations while enhancing evaluation diversity and fidelity.

Contribution

It presents a novel multi-level action-conditioning mechanism and ray map encoding for dynamic multi-view image generation in robotic environments.

Findings

Effective generation of realistic, action-conditioned visual observations.

Enhanced generalization through diverse failure trajectory training.

Significant cost reduction in robotic evaluation processes.

Abstract

Robotic imitation learning has advanced from solving static tasks to addressing dynamic interaction scenarios, but testing and evaluation remain costly and challenging due to the need for real-time interaction with dynamic environments. We propose EnerVerse-AC (EVAC), an action-conditional world model that generates future visual observations based on an agent's predicted actions, enabling realistic and controllable robotic inference. Building on prior architectures, EVAC introduces a multi-level action-conditioning mechanism and ray map encoding for dynamic multi-view image generation while expanding training data with diverse failure trajectories to improve generalization. As both a data engine and evaluator, EVAC augments human-collected trajectories into diverse datasets and generates realistic, action-conditioned video observations for policy testing, eliminating the need for physical robots or complex simulations. This approach significantly reduces costs while maintaining high fidelity in robotic manipulation evaluation. Extensive experiments validate the effectiveness of our method. Code, checkpoints, and datasets can be found at <https://annaj2178.github.io/EnerverseAC.github.io>.