mlsanityData to 14 Oct 2026

Generative Trajectory Stitching through Diffusion Composition

arXiv 2503.05153 ↗First posted 7 Mar 2025cs.RO cs.AI cs.LGcited by 41

On the map: Diffusion models → Diffusion and flow policies for robotics

Abstract (arXiv)

Effective trajectory stitching for long-horizon planning is a significant challenge in robotic decision-making. While diffusion models have shown promise in planning, they are limited to solving tasks similar to those seen in their training data. We propose CompDiffuser, a novel generative approach that can solve new tasks by learning to compositionally stitch together shorter trajectory chunks from previously seen tasks. Our key insight is modeling the trajectory distribution by subdividing it into overlapping chunks and learning their conditional relationships through a single bidirectional diffusion model. This allows information to propagate between segments during generation, ensuring physically consistent connections. We conduct experiments on benchmark tasks of various difficulties, covering different environment sizes, agent state dimension, trajectory types, training data quality, and show that CompDiffuser significantly outperforms existing methods.

Key facts. Cited by 41 papers in the map (data through 14 Oct 2026). Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models. Builds on 81 of its 91 references that are in the map. First posted on arXiv 7 Mar 2025.

Builds on

81 of 91 references.

Closest ancestors · 26

Same direction, newest first

  1. 2025What Makes a Good Diffusion Planner for Decision Making?1 Mar
  2. 2025DRDT3: Diffusion-Refined Decision Test-Time Training Model12 Jan
  3. 2024Causal Composition Diffusion Model for Closed-loop Traffic Generation23 Dec
  4. 2024DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving22 Nov
  5. 2024Enhancing Decision Transformer with Diffusion-Based Trajectory Branch Generation18 Nov
  6. 2024Composing Diffusion Policies for Few-shot Learning of Movement Trajectories22 Oct
  7. 2024Diffusion Model for Planning: A Systematic Literature Review16 Aug
  8. 2024Hierarchical Diffusion Policy for Kinematics-Aware Multi-Task Robotic Manipulation6 Mar
  9. 2024Stitching Sub-Trajectories with Conditional Diffusion Model for Goal-Conditioned Offline RL11 Feb
  10. 2024Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous Driving and Zero-Shot Instruction Following9 Feb
  11. 2024Simple Hierarchical Planning with Diffusion5 Jan
  12. 2023Refining Diffusion Planner for Reliable Behavior Synthesis by Automatic Detection of Infeasible Plans30 Oct
14 more
  1. 2023Adaptive Online Replanning with Diffusion Models14 Oct
  2. 2023Generative Skill Chaining: Long-Horizon Skill Planning with Diffusion Models13 Oct
  3. 2023AlignDiff: Aligning Diverse Human Preferences via Behavior-Customisable Diffusion Model3 Oct
  4. 2023Reasoning with Latent Diffusion in Offline Reinforcement Learning12 Sep
  5. 2023Motion Planning Diffusion: Learning and Planning of Robot Motions with Diffusion Models3 Aug
  6. 2023DiMSam: Diffusion Models as Samplers for Task and Motion Planning under Partial Observability22 Jun
  7. 2023Extracting Reward Functions from Diffusion Models1 Jun
  8. 2023Diffusion Model is an Effective Planner and Data Synthesizer for Multi-Task Reinforcement Learning29 May
  9. 2023MADiff: Offline Multi-agent Learning with Diffusion Models27 May
  10. 2023EDGI: Equivariant Diffusion for Planning with Embodied Agents22 Mar
  11. 2023AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Planners3 Feb
  12. 2023Imitating Human Behaviour with Diffusion Models25 Jan
  13. 2022Is Conditional Generative Modeling all you need for Decision-Making?28 Nov
  14. 2022Diffusion Policies as an Expressive Policy Class for Offline Reinforcement Learning12 Aug

Other references · 55

  1. 2017Attention Is All You Needcited by 60,361
  2. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  3. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,058
  4. 2022Scalable Diffusion Models with Transformerscited by 6,054
  5. 2021Decision Transformer: Reinforcement Learning via Sequence Modelingcited by 1,546
  6. 2021Offline Reinforcement Learning with Implicit Q-Learningcited by 1,293
  7. 2022Planning with Diffusion for Flexible Behavior Synthesiscited by 1,061
  8. 2024Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusioncited by 681
  9. 2022Compositional Visual Generation with Composable Diffusion Modelscited by 624
  10. 2023Human Motion Diffusion as a Generative Priorcited by 357
  11. 2019Learning Latent Plans from Playcited by 316
  12. 2022Contrastive Learning as Goal-Conditioned Reinforcement Learningcited by 253
38 more
  1. 2024OGBench: Benchmarking Offline Goal-Conditioned RLcited by 238
  2. 2023Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMCcited by 220
  3. 2021RvS: What is Essential for Offline RL via Supervised Learning?cited by 185
  4. 2023HIQL: Offline Goal-Conditioned RL with Latent States as Actionscited by 141
  5. 2024Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planningcited by 135
  6. 2023Optimal Goal-Reaching Reinforcement Learning via Quasimetric Learningcited by 121
  7. 2022Q-learning Decision Transformer: Leveraging Dynamic Programming for Conditional Sequence Modelling in Offline RLcited by 103
  8. 2023DiffCollage: Parallel Generation of Large Content with Diffusion Modelscited by 103
  9. 2023Synthetic Experience Replaycited by 101
  10. 2023Compositional Foundation Models for Hierarchical Planningcited by 95
  11. 2023Compositional Abilities Emerge Multiplicatively: Exploring Diffusion Models on a Synthetic Taskcited by 93
  12. 2023Elastic Decision Transformercited by 67
  13. 2024PoCo: Policy Composition from and for Heterogeneous Robot Learningcited by 42
  14. 2024Compositional Generative Modeling: A Single Model is Not All You Needcited by 41
  15. 2024Potential Based Diffusion Motion Planningcited by 41
  16. 2023Compositional Diffusion-Based Continuous Constraint Solverscited by 40
  17. 2023PDPP: Projected Diffusion for Procedure Planning in Instructional Videoscited by 36
  18. 2024DiffStitch: Boosting Offline Reinforcement Learning with Diffusion-based Trajectory Stitchingcited by 36
  19. 2024The Superposition of Diffusion Models Using the Ito Density Estimatorcited by 36
  20. 2024EnergyMoGen: Compositional Human Motion Generation with Energy-Based Diffusion Model in Latent Spacecited by 30
  21. 2023Probabilistic Adaptation of Text-to-Video Modelscited by 29
  22. 2023Compositional Sculpting of Iterative Generative Processescited by 27
  23. 2024Closing the Gap between TD Learning and Supervised Learning -- A Generalisation Point of Viewcited by 27
  24. 2023Critic-Guided Decision Transformer for Offline Reinforcement Learningcited by 26
  25. 2023ACT: Empowering Decision Transformer with Dynamic Programming via Advantage Conditioningcited by 25
  26. 2024Reinformer: Max-Return Sequence Modeling for Offline RLcited by 25
  27. 2024GTA: Generative Trajectory Augmentation with Guidance for Offline Reinforcement Learningcited by 22
  28. 2025Composition and Control with Distilled Energy Diffusion Models and Sequential Monte Carlocited by 20
  29. 2024CoMA: Compositional Human Motion Generation with Multi-modal Agentscited by 19
  30. 2022Model-based Trajectory Stitching for Improved Offline Reinforcement Learningcited by 15
  31. 2024Generative Factor Chaining: Coordinated Manipulation with Diffusion-based Factor Graphcited by 14
  32. 2025Mechanisms of Projective Composition of Diffusion Modelscited by 14
  33. 2023Free from Bellman Completeness: Trajectory Stitching via Model-based Return-conditioned Supervised Learningcited by 13
  34. 2024Resisting Stochastic Risks in Diffusion Planners with the Trajectory Aggregation Treecited by 13
  35. 2024Compositional Risk Minimizationcited by 13
  36. 2024Compositional Image Decomposition with Diffusion Modelscited by 12
  37. 2022BATS: Best Action Trajectory Stitchingcited by 10
  38. 2023Goal-Conditioned Predictive Coding for Offline Reinforcement Learningcited by 10

50 most cited of 55 All 55 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 18; 2026: 23.

41 papers.

Sort:

By area:

  1. 2025Reinforcement Learning for Flow-Matching Policiescited by 15
  2. 2025State-Covering Trajectory Stitching for Diffusion Plannerscited by 13
  3. 2025Multi-Modal Manipulation via Multi-Modal Policy Consensuscited by 13
  4. 2025Self-Predictive Representations for Combinatorial Generalization in Behavioral Cloningcited by 12
  5. 2025CREPE: Controlling Diffusion with Replica Exchangecited by 11
  6. 2025Local Manifold Approximation and Projection for Manifold-Aware Diffusion Planningcited by 9
  7. 2025Test-Time Graph Search for Goal-Conditioned Reinforcement Learningcited by 8
  8. 2025Hybrid Diffusion for Simultaneous Symbolic and Continuous Planningcited by 7
  9. 2026Compositional Planning with Jumpy World Modelscited by 7
  10. 2026Chain-of-Goals Hierarchical Policy for Long-Horizon Offline Goal-Conditioned RLcited by 5
  11. 2026Diffusion ReRoll: Revisable Denoising for Robotic Sequential Predictioncited by 5
  12. 2025BiTrajDiff: Bidirectional Trajectory Generation with Diffusion Models for Offline Reinforcement Learningcited by 4
29 more
  1. 2025Generative View Stitchingcited by 4
  2. 2026Temporal Backtracking Search for Test-time Generative Video Reasoningcited by 4
  3. 2025Extremum Flow Matching for Offline Goal Conditioned Reinforcement Learningcited by 3
  4. 2025Factorizing Diffusion Policies for Observation Modality Prioritizationcited by 3
  5. 2025Model-Based Diffusion Sampling for Predictive Control in Offline Decision Makingcited by 3
  6. 2025Compositional Monte Carlo Tree Diffusion for Extendable Planningcited by 2
  7. 2026Self-Imitated Diffusion Policy for Efficient and Robust Visual Navigationcited by 2
  8. 2026Compositional Visual Planning via Inference-Time Diffusion Scalingcited by 2
  9. 2026Refining Compositional Diffusion for Reliable Long-Horizon Planningcited by 2
  10. 2026NFTR: From Provable Mode-Averaging to Geodesic Subgoal Selection in Offline Goal-Conditioned RLcited by 2
  11. 2025Flexible Locomotion Learning with Diffusion Model Predictive Controlcited by 1
  12. 2025ASTRO: Adaptive Stitching via Dynamics-Guided Trajectory Rolloutscited by 1
  13. 2026Action-Sufficient Goal Representationscited by 1
  14. 2026Analogical Trajectory Transfercited by 1
  15. 2026Compositional Transduction with Latent Analogies for Offline Goal-Conditioned Reinforcement Learningcited by 1
  16. 2025Improving Trajectory Stitching with Flow Modelscited by 0
  17. 2025What Do You Need for Compositional Generalization in Diffusion Planning?cited by 0
  18. 2026Muninn: Your Trajectory Diffusion Model But Fastercited by 0
  19. 2026Bridging Domain Gaps with Target-Aligned Generation for Offline Reinforcement Learningcited by 0
  20. 2026Plan First, Diffuse Later: Extrinsic Graph Guidance for Long-Horizon Diffusion Planningcited by 0
  21. 2026Sparse Compositional Flow Matching by geometric assembly from motion primitivescited by 0
  22. 2026Coarse-to-Fine Compositional Diffusion for Long-Horizon Planningcited by 0
  23. 2026ChronoForest: Closed-Loop Multi-Tree Diffusion Planning for Efficient Bridge Search and Route Compositioncited by 0
  24. 2026Compositional Generative Modeling from Decentralized Datacited by 0
  25. 2026Human2Any: Human-to-Robot Transfer via Constraint-Aware Compositional Planningcited by 0
  26. 2026HorizonFlow: Variable-Length Planning for Offline Goal-Conditioned RLcited by 0
  27. 2026NEEDLEWORK: Offline Rewriting of Robot Data with Verified Local Stitchescited by 0
  28. 2026World Action Modeling with Progressive Visual Planningcited by 0
  29. 2026Imagining Goals with Geometric Horizon Models for Policy Generalizationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Cascaded Diffusion Models for Neural Motion Planning
  2. 2026Amortizing Trajectory Diffusion with Keyed Drift Fields
  3. 2025STITCH-OPE: Trajectory Stitching with Guided Diffusion for Off-Policy Evaluation
  4. 2024Diffusion Meets Options: Hierarchical Generative Skill Composition for Temporally-Extended Tasks
  5. 2024Motion Planning Diffusion: Learning and Adapting Robot Motion Planning with Diffusion Models
  6. 2024LDP: A Local Diffusion Planner for Efficient Robot Navigation and Collision Avoidance
  7. 2026Model-Based Online Decision Making via Generative Trajectory Planning
  8. 2024Multi-Robot Motion Planning with Diffusion Models