Vista helps robots actively explore to improve task success
VISTA: Active Exploration for Self-Improvement in Embodied Agents
Robotics
Summary
Robots that interact with the physical world often learn from past attempts, but traditionally they only analyze old experiences without trying new actions to fix mistakes. The authors propose VISTA, a system where a helpful 'meta-agent' makes robots test and explore actively to improve their problem-solving strategies. VISTA keeps track of past strategies and their results to avoid repeating failures and to build on successes. This approach leads to robots completing tasks more successfully using fewer trial attempts.
What this means in practice
- •For robotic system developers: Improve robot task success by enabling active exploration and iterative strategy refinement based on execution feedback.
- •For simulation platform engineers: Reduce the number of simulator rollouts required for embodied agents to refine behaviors by using contingent self-exploration.
Authors
Zhang Zhang, Shengjun Zhang, Simin Huang, Min Chen, Chong Xia, Xin Fei, Yueqi Duan
Abstract
Agentic embodied intelligence plays a foundational role in open-ended, dynamic physical environments. Rather than treating policy learning as an end-to-end mapping from instructions to actions, recent studies reveal that embodied systems depend not only on pre-trained foundation models, but also on harness. Prior approaches largely treat rollouts as static logs for post-hoc reflection, producing harness updates without autonomously testing and refining them through subsequent embodied interaction. To address this challenge, we propose VISTA, a meta-agent framework that transforms post-hoc rollout reflection into active self-exploration for embodied harness improvement. Specifically, VISTA maintains a harness graph that records the lineage of proposed harnesses, which enables the meta-agent to retrieve prior hypotheses, build on effective mechanisms, and avoid repeatedly exploring failed directions. Each iteration begins from a retained harness version and its execution evidence. Rather than directly committing an update from these historical trajectories, the meta-agent actively explores in the environment. It autonomously generates new rollouts, probes execution failures and recovery behaviors, and iteratively modifies the harness based on the resulting evidence. The refined harness is then evaluated afresh and added to the harness graph. We evaluate VISTA on the LIBERO-Pro, RoboCasa, and RoboTwin benchmarks. Experimental results demonstrate that VISTA progressively improves task success and is more rollout-efficient than static reflection, achieving higher success rates with fewer simulator rollouts.