Robot uses past mistakes to improve complex multi-step tasks
ReTrace: Cross-Stage Failure Attribution for Self-Improving Robotic Manipulation
Robotics
Summary
Robots often struggle with tasks that have many steps because a problem at one step can cause failure later on. The authors created ReTrace, a system that helps robots learn from their mistakes by tracing failures back to the original wrong moves and trying better actions. ReTrace does this without needing example demonstrations or changes to the robot’s underlying model. Tested on real-world tasks, the robot using ReTrace greatly improved success in tricky tasks like plug insertion and ball pouring.
What this means in practice
- •For robotic system integrators: Enable robots to independently identify and fix early mistakes that cause later failures in complex manipulation tasks, improving overall success rates.
- •For industrial automation teams: Deploy robotic arms that adapt physical task skills over time without requiring new demonstrations or retraining, reducing manual tuning costs.$Commercial implications: Makes it possible to sell adaptable robotic solutions that improve performance autonomously, reducing downtime and engineering effort in manufacturing.
Authors
Yue Yu, Bin Zhu, Junhui Li, Jiayu Wang, Shijie Zhou, Jingjing Chen
Abstract
General-purpose multimodal foundation models are increasingly capable of directly controlling robots, yet improving their manipulation capabilities through autonomous interaction remains challenging. A key challenge to such self-improvement is learning from failures in complex manipulation tasks involving cross-stage dependencies, where an action may succeed locally but cause a later stage to fail. To address this challenge, we introduce ReTrace, a self-improvement framework for robotic manipulation through autonomous physical interaction and retrospective reasoning, without provided demonstrations or model-parameter updates. ReTrace retrospectively traces downstream failures to earlier decisions, proposes targeted corrections, and evaluates them through subsequent interactions. Across attempts, ReTrace progressively consolidates validated corrections into a reusable manipulation skill that guide subsequent execution. We evaluate ReTrace on four real-world manipulation tasks using a Franka Research 3 robot. On the two tasks with strong cross-stage dependencies, GPT-6 Astra-low improves from 0% to 80% success on plug insertion and from 0% to 100% on ball pouring, while GPT-6 Sol-xhigh improves from 0% to 10% and 0% to 90%, respectively. The learned skills also transfer across model and reasoning configurations and reduce inference-time decision cost in most evaluated settings.