graph-guided planning

Handcrafted heuristic functions in search-based navigation suffer from poor generalization across unseen maps and long-distance paths. Method: This paper proposes a local heuristic learning framework that explicitly defines and end-to-end learns either heuristic bias correction or local cost estimation within a spatial neighborhood—replacing conventional global heuristic modeling. By decomposing complex global prediction into lightweight local regression, the approach significantly reduces learning complexity. Integrated with graph search algorithms (e.g., A*), it operates under supervised learning using local state inputs while preserving bounded suboptimality guarantees. Contribution/Results: Experiments demonstrate 2–20× reduction in node expansions, improved training efficiency, and robust generalization to both unseen maps and long-range trajectories—without compromising solution quality or theoretical guarantees.

Classical planning suffers from inefficient search-space utilization and poor action selection due to the decoupling of learning and search. Method: This paper proposes a novel partial-space search paradigm: (i) it models action dependencies via PDDL’s relational structure to define fine-grained action subspaces for early action pruning; (ii) it introduces an action-set heuristic that supports both automatic conversion from symbolic heuristics and end-to-end training, enabling bidirectional alignment between search structure and learned signals; and (iii) it implements the LazyLifted planner architecture, integrating supervised learning on large-scale search trajectories. Results: On the IPC 2023 Learning Track, it significantly outperforms state-of-the-art ML-based heuristics, demonstrating superior efficiency and robustness—especially in high-branching-factor domains—and achieves overall performance exceeding that of LAMA.

Motion planning for high-dimensional robotic systems—such as manipulator arms and mobile manipulation platforms—often suffers from high computational cost and solution instability. This work proposes the Multi-Graph Search (MGS) algorithm, which introduces, for the first time, a multi-implicit-graph structure into search-based motion planning. By concurrently maintaining and incrementally expanding multiple state subgraphs, MGS dynamically focuses computational effort on high-potential regions and merges subgraphs during search to efficiently discover feasible paths. The method is theoretically guaranteed to be complete and bounded-suboptimal. Experimental results demonstrate that MGS significantly outperforms existing approaches across a variety of high-dimensional tasks, achieving notable advances in computational efficiency, solution consistency, and scalability.

To address state-set storage explosion and excessive memory overhead in large-scale explicit-state-space search, this paper proposes the Dynamic Trie Database (DTDB)—the first trie-based structure extended from static to fully dynamic, supporting incremental insertions and deletions without pre-allocated memory. DTDB guarantees theoretically provable compression ratios while maintaining high query efficiency. It uniformly compresses states encoded with both propositional and numeric variables, and is applicable to both grounded and lifted settings in classical and numeric planning. Experimental evaluation across multiple benchmark domains demonstrates memory compression ratios of 10×–1000× over baseline methods, with runtime overhead under 2%. This substantially improves scalability in representing and processing states for large-scale planning problems.

This work addresses the bottleneck in hierarchical planning for continuous-state/action domains—namely, its reliance on manually engineered symbolic predicates for state abstraction. We propose the first fully automated predicate invention framework. Methodologically, it employs grammar-guided differentiable search over predicate sets, jointly optimizing high-level operators and low-level samplers via proxy objective optimization and demonstration-guided symbolic predicate learning—enabling end-to-end abstraction discovery and hierarchical planning co-training. Evaluated on four robotic planning benchmarks, our approach significantly outperforms six baselines, demonstrating strong generalization: it rapidly solves unseen tasks. Our core contributions are twofold: (1) the first fully automated, human-intervention-free predicate invention mechanism; and (2) empirical validation that automatically discovered predicates substantially improve both planning efficiency and cross-task generalization.

This work addresses the exponential growth in computational complexity with planning horizon that plagues online planning in continuous Markov decision processes due to tree-based structures. To overcome this limitation, the authors propose the Graph-based Sparse Sampling (GSS) algorithm, which introduces a branching-free graph structure into online planning for continuous MDPs for the first time. GSS enables efficient allocation of computational resources by sharing sampled future trajectories across multiple candidate actions and integrating smooth backtracking with heuristic policies, while also supporting GPU-accelerated batch processing. Under conditions of trajectory overlap, regularity, and action coverage, theoretical analysis shows that GSS incurs only polynomial growth in performance error with respect to the planning horizon, thereby circumventing the exponential bottleneck inherent in traditional tree search. Empirical results demonstrate that GSS significantly outperforms existing tree-based planners in continuous control tasks, achieving near-optimal performance especially in long-horizon scenarios.

This work addresses the challenge of maintaining global trajectory consistency in long-horizon planning tasks, where diffusion models often struggle due to their reliance on local denoising steps. To overcome this limitation, the authors propose the eXtrinsic search-guided Diffuser (XDiffuser) framework, which introduces an extrinsic graph-based search mechanism for the first time. Specifically, a lightweight path planner operates on a state-space graph to generate a coarse trajectory, which then guides the diffusion model to produce a complete trajectory in a single denoising pass. This design shifts the exploration burden from the computationally intensive diffusion process to an efficient graph search, significantly improving both planning efficiency and generalization. Empirical results demonstrate that XDiffuser outperforms existing diffusion-based baselines in multi-agent coordination and TSP-like reasoning tasks, with particularly strong performance under low-quality training data and unseen compositional scenarios.

This work addresses the poor generalization and inefficient exploration of deep reinforcement learning (DRL) in sparse-reward combinatorial planning tasks by proposing a self-improving weighted A* (WA*) learning framework. Integrating relational graph neural networks with symbolic state representations, the method employs Q-learning to self-supervise the update of its heuristic function, forming a closed loop with the search process—without requiring expert demonstrations or counterfactual relabeling. The approach achieves, for the first time, strong zero-shot generalization: it substantially outperforms conventional DRL methods on benchmarks including Sokoban, PushWorld, The Witness, and IPC-2023, and successfully transfers policies trained on small-scale Blocksworld instances (30 blocks) to solve vastly larger ones (488 blocks) without further training.

This work addresses the challenge of jointly planning spatiotemporal constraints—specifying when and where tasks are executed—and topological constraints—governing agent interaction structures—in multi-agent systems. To this end, the authors propose a unified modeling framework grounded in STL-GO logic, which for the first time incorporates dynamic multi-graph interactions. They develop two complete solution approaches based on Mixed-Integer Programming (MIP) and Satisfiability Modulo Theories (SMT), enabling seamless switching and comparative analysis between the two paradigms. The effectiveness of the proposed method is validated on a multi-UAV search-and-rescue benchmark, demonstrating strong expressiveness and favorable scalability across varying team sizes and levels of time-varying graph complexity.