Towards Reliable Code-as-Policies: A Neuro-Symbolic Framework for Embodied Task Planning

TL;DR

The paper tackles the reliability gap of LLM-generated code-as-policies in embodied robotics under dynamic, partially observable conditions. It introduces NeSyRo, a neuro-symbolic framework that interleaves symbolic verification with interactive validation to produce grounded, executable policy code, leveraging safe exploratory probes when observations are missing. Empirical results in RLBench and real-world tasks show NeSyRo substantially improves task success rates (e.g., 46.2% over Code-as-Policies baselines) and achieves high executability of task-relevant actions (over 86.8%), validating robustness against uncertainty. By combining Commonsense and Logic-based confidence within a recursive verification-validation loop, NeSyRo enhances reliability and safety of robotic task planning in uncertain environments, with broad implications for real-world embodied AI.

Abstract

Recent advances in large language models (LLMs) have enabled the automatic generation of executable code for task planning and control in embodied agents such as robots, demonstrating the potential of LLM-based embodied intelligence. However, these LLM-based code-as-policies approaches often suffer from limited environmental grounding, particularly in dynamic or partially observable settings, leading to suboptimal task success rates due to incorrect or incomplete code generation. In this work, we propose a neuro-symbolic embodied task planning framework that incorporates explicit symbolic verification and interactive validation processes during code generation. In the validation phase, the framework generates exploratory code that actively interacts with the environment to acquire missing observations while preserving task-relevant states. This integrated process enhances the grounding of generated code, resulting in improved task reliability and success rates in complex environments. We evaluate our framework on RLBench and in real-world settings across dynamic, partially observable scenarios. Experimental results demonstrate that our framework improves task success rates by 46.2% over Code-as-Policies baselines and attains over 86.8% executability of task-relevant actions, thereby enhancing the reliability of task planning in dynamic environments.

Figures (10)

• Figure 1: Concept of our $\textsc{NeSyRo}\xspace$ framework illustrated with an example of a room-cleaning task where drawer states are initially unknown. While (1) naive code generation fails without detecting missing observations, (2) $\textsc{NeSyRo}\xspace$ recursively probes the environment to recover drawer states, enabling the generation of grounded code that successfully completes the task.
• Figure 2: The $\textsc{NeSyRo}\xspace$ framework with Neuro-Symbolic Code Verification and Neuro-Symbolic Code Validation phases. It recursively verifies and validates the policy code, while incrementally acquiring observations.
• Figure 3: Representative failure scenarios under partial observability across real-world tasks
• Figure 4: Real-world compile error rate over all task types
• Figure 5: Real-world example of safe probe and policy code refinement in partially observable setting