Toward Deployable Multi-Robot Collaboration via a Symbolically-Guided Decision Transformer

2025-08-19

AI Summary

Reinforcement learning for multi-robot collaborative manipulation suffers from high data demand and strict Markovian assumptions, while existing Decision Transformers (DTs) lack hierarchical modeling and interpretability. Method: We propose a symbol-guided hierarchical decision framework—the first to integrate DTs into multi-robot manipulation. It employs a neuro-symbolic planner to generate interpretable symbolic subgoals for high-level semantic planning, and a goal-conditioned Decision Transformer (GCDT) to model long-horizon action sequences at the low level. The framework unifies causal Transformer architectures with neuro-symbolic reasoning to enable zero-shot and few-shot cross-task transfer. Contribution/Results: Experiments demonstrate significant improvements in deployment efficiency and generalization capability of multi-robot systems operating in complex, dynamic environments, validating the framework’s scalability, interpretability, and sample efficiency.

Technology Category

• Intelligent Robots: Multi-Robot Systems
• Machine Learning: Neuro-Symbolic Learning
• Multiagent Systems: Multiagent Planning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Reinforcement learning (RL) has demonstrated great potential in robotic operations. However, its data-intensive nature and reliance on the Markov Decision Process (MDP) assumption limit its practical deployment in real-world scenarios involving complex dynamics and long-term temporal dependencies, such as multi-robot manipulation. Decision Transformers (DTs) have emerged as a promising offline alternative by leveraging causal transformers for sequence modeling in RL tasks. However, their applications to multi-robot manipulations still remain underexplored. To address this gap, we propose a novel framework, Symbolically-Guided Decision Transformer (SGDT), which integrates a neuro-symbolic mechanism with a causal transformer to enable deployable multi-robot collaboration. In the proposed SGDT framework, a neuro-symbolic planner generates a high-level task-oriented plan composed of symbolic subgoals. Guided by these subgoals, a goal-conditioned decision transformer (GCDT) performs low-level sequential decision-making for multi-robot manipulation. This hierarchical architecture enables structured, interpretable, and generalizable decision making in complex multi-robot collaboration tasks. We evaluate the performance of SGDT across a range of task scenarios, including zero-shot and few-shot scenarios. To our knowledge, this is the first work to explore DT-based technology for multi-robot manipulation.

Problem

• Enabling deployable multi-robot collaboration through symbolic guidance
• Addressing limitations of reinforcement learning in complex robotic operations
• Integrating neuro-symbolic planning with transformer-based decision making

Innovation

• Integrates neuro-symbolic planner with transformer
• Generates symbolic subgoals for hierarchical guidance
• Uses goal-conditioned transformer for robot decisions