High-level robot plans can miss hidden safety problems during execution
Mind the Refinement Gap: When Safe High-Level Robot Plans Produce Unsafe Executions
Machine LearningRobotics
Summary
Robots often plan their actions using simple, high-level steps that are checked for safety before making a move. This paper shows that these safety checks can miss problems because the actual detailed steps robots take might cause unexpected unsafe situations. The authors studied a robot system called RoboGuard and found that checking the detailed sequence of actions helps catch these hidden safety issues. They propose a method called graph-based trace refinement to improve safety monitoring during robot operation.
What this means in practice
- •For robotics software engineers: Implement graph-based trace refinement to improve safety checks on robot action sequences, reducing risks from overlooked details in high-level plans.
- •For natural language interface developers: Use scene-grounded safety specifications to better align robot plans generated from natural language with safe real-world executions.
Authors
Stabak Das, Priyesh Ranjan, Xiangfang Li, Lijun Qian
Abstract
Language-enabled robot systems increasingly combine semantic-graph planning with temporal-logic safety monitors. We investigate a trace-completeness assumption in these systems: whether the high-level action sequence checked by a monitor represents the navigation and implicit action effects induced during execution. We audit this assumption in RoboGuard by comparing its verdict on a surface plan with its verdict on a graph-refined trace under the same Linear Temporal Logic (LTL) specification. Our evaluation comprises 28 controlled cases spanning five action-abstraction families and 14 end-to-end cases in which SPINE [1] generates plans from natural-language instructions while RoboGuard generates scene-grounded safety specifications. In the controlled evaluation, all 12 targeted abstraction cases exhibit the predicted surface-versus-refined discrepancy while all 16 controls behave as expected, motivating graph-based trace refinement as a lightweight mitigation and a diagnostic tool for physical-AI safety monitors.