SDA-PLANNER: State-Dependency Aware Adaptive Planner for Embodied Task Planning

2025-09-30

AI Summary

Existing LLM-based embodied task planning approaches suffer from three key limitations: rigid planning paradigms, insufficient action constraints, and weak error recovery capabilities. This paper proposes a state-dependency-aware adaptive planning framework. Our method integrates LLM-driven task decomposition, state-dependency graph reasoning, and execution feedback optimization. Specifically, we (1) construct an explicit state-dependency graph modeling action preconditions and postconditions to support dynamic environment representation; (2) design an error-driven backtracking diagnosis and local subtree reconstruction mechanism, enabling adaptive switching of planning paradigms under closed-loop feedback; and (3) unify symbolic reasoning with neural generation for robust plan refinement. Experiments across diverse error-prone scenarios demonstrate significant improvements in task success rate and goal completion, validating the framework’s strong robustness and cross-task generalization capability.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Planning with Language Models
• Multiagent Systems: Multiagent Planning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Embodied task planning requires agents to produce executable actions in a close-loop manner within the environment. With progressively improving capabilities of LLMs in task decomposition, planning, and generalization, current embodied task planning methods adopt LLM-based architecture.However, existing LLM-based planners remain limited in three aspects, i.e., fixed planning paradigms, lack of action sequence constraints, and error-agnostic. In this work, we propose SDA-PLANNER, enabling an adaptive planning paradigm, state-dependency aware and error-aware mechanisms for comprehensive embodied task planning. Specifically, SDA-PLANNER introduces a State-Dependency Graph to explicitly model action preconditions and effects, guiding the dynamic revision. To handle execution error, it employs an error-adaptive replanning strategy consisting of Error Backtrack and Diagnosis and Adaptive Action SubTree Generation, which locally reconstructs the affected portion of the plan based on the current environment state. Experiments demonstrate that SDA-PLANNER consistently outperforms baselines in success rate and goal completion, particularly under diverse error conditions.

Problem

• Addresses fixed planning paradigms in embodied task planning
• Introduces state-dependency aware mechanisms for action constraints
• Implements error-adaptive replanning strategies for execution failures

Innovation

• Adaptive planning paradigm for embodied task execution
• State-Dependency Graph models action preconditions and effects
• Error-adaptive replanning strategy with local reconstruction