Robot learning improves with safety guards to handle hard tasks
Learning Beyond What Humans Can Demonstrate
Robotics
Summary
Some robot tasks are too tricky for humans to show how to do them directly, like balancing or precise timing tasks. The authors developed GLIDE, a system that learns where mistakes happen and creates rules to avoid those mistakes during robot training and use. GLIDE helps robots get better data and safer actions even without perfect human demonstrations. This approach improved success rates dramatically on challenging tasks like moving a plate, handing over a marker, and serving wine.
What this means in practice
- •For robotics engineers: Improve training data and policy safety for complex robot manipulation tasks without requiring perfect human demonstrations.
- •For industrial automation teams: Safely deploy robot policies in manufacturing tasks that involve precise timing and contact by filtering risky actions automatically.
Authors
Yuchen Song, Aditya Mittal, Unnat Jain
Abstract
Behavior cloning for robot manipulation relies on expert demonstrations. However, for tasks that require dynamic stability, precise contact timing, or dexterous coordination, human operators may find it hard or even impossible to collect data. We study this infeasible-demonstration regime and propose GLIDE: Guardrails for Learning from Infeasible Demonstrations Efficiently, a framework that infers task-specific failure modes and converts them into executable guardrails for data collection and policy deployment. Given a task description and the conditioning teleoperation code, GLIDE writes guardrails that use system states to filter teleoperation and policy commands, constrain failure-prone actions, and iteratively improve from trajectory feedback. Across three tasks, GLIDE discovers emergent guardrails that go beyond domain-expert hardcoded ones, improving data collection over naive VR teleoperation and domain-expert hardcoded guardrails. After refinement, GLIDE raises data-collection success from 0-10 percent to 70-90 percent across the three tasks. During policy execution, mixed-data guarded policies reach 70 percent, 60 percent, and 60 percent success on Tomato plate transfer, Marker handover and stand, and Wine serving tasks. These results show that GLIDE can support policy learning when direct demonstrations are infeasible. Project website: http://guardrail-policy.github.io/