mlsanityData to 14 Oct 2026

Learning to Ball: Composing Policies for Long-Horizon Basketball Moves

arXiv 2509.22442 ↗First posted 26 Sep 2025cs.GR cs.AI cs.LGcited by 6

On the map: Deep reinforcement learning methods → RL for robot locomotion and manipulation

Abstract (arXiv)

Learning a control policy for a multi-phase, long-horizon task, such as basketball maneuvers, remains challenging for reinforcement learning approaches due to the need for seamless policy composition and transitions between skills. A long-horizon task typically consists of distinct subtasks with well-defined goals, separated by transitional subtasks with unclear goals but critical to the success of the entire task. Existing methods like the mixture of experts and skill chaining struggle with tasks where individual policies do not share significant commonly explored states or lack well-defined initial and terminal states between different phases. In this paper, we introduce a novel policy integration framework to enable the composition of drastically different motor skills in multi-phase long-horizon tasks with ill-defined intermediate states. Based on that, we further introduce a high-level soft router to enable seamless and robust transitions between the subtasks. We evaluate our framework on a set of fundamental basketball skills and challenging transitions. Policies trained by our approach can effectively control the simulated character to interact with the ball and accomplish the long-horizon task specified by real-time user commands, without relying on ball trajectory references.

Key facts. Cited by 6 papers in the map (data through 14 Oct 2026). Placed in RL for robot locomotion and manipulation, one of 15 directions in Deep reinforcement learning methods. Builds on 43 of its 80 references that are in the map. First posted on arXiv 26 Sep 2025.

Builds on

43 of 80 references.

References · 43

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2017Proximal Policy Optimization Algorithmscited by 15,215
  3. 2017Improved Training of Wasserstein GANscited by 2,235
  4. 2019AMASS: Archive of Motion Capture as Surface Shapescited by 1,236
  5. 2021Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learningcited by 998
  6. 2016The Option-Critic Architecturecited by 448
  7. 2020GRAB: A Dataset of Whole-Body Human Grasping of Objectscited by 405
  8. 2017FeUdal Networks for Hierarchical Reinforcement Learningcited by 337
  9. 2021DexYCB: A Benchmark for Capturing Hand Grasping of Objectscited by 325
  10. 2016Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivationcited by 286
  11. 2017Geometric GANcited by 233
  12. 2021Character Controllers Using Motion VAEscited by 184
31 more
  1. 2018DeepMimic: Example-Guided Deep Reinforcement Learning of Physics-Based Character Skillscited by 163
  2. 2024DexCap: Scalable and Portable Mocap Data Collection System for Dexterous Manipulationcited by 160
  3. 2024TRAM: Global Trajectory and Motion of 3D Humans from in-the-wild Videoscited by 106
  4. 2023CALM: Conditional Adversarial Latent Models for Directable Virtual Characterscited by 94
  5. 2023Unified Human-Scene Interaction via Prompted Chain-of-Contactscited by 94
  6. 2016A Deep Hierarchical Approach to Lifelong Learning in Minecraftcited by 90
  7. 2024Expressive Whole-Body 3D Gaussian Avatarcited by 89
  8. 2021From Motor Control to Team Play in Simulated Humanoid Footballcited by 87
  9. 2016Learning values across many orders of magnitudecited by 85
  10. 2017Learning human behaviors from motion capture by adversarial imitationcited by 77
  11. 2023PhysHOI: Physics-Based Imitation of Dynamic Human-Object Interactioncited by 76
  12. 2023MoConVQ: Unified Physics-Based Motion Control via Scalable Discrete Representationscited by 60
  13. 2023C·ASE: Learning Conditional Adversarial Skill Embeddings for Physics-based Characterscited by 55
  14. 2020ALLSTEPS: Curriculum-driven Learning of Stepping Stone Skillscited by 54
  15. 2023Sequential Dexterity: Chaining Dexterous Policies for Long-Horizon Manipulationcited by 49
  16. 2021Adversarial Skill Chaining for Long-Horizon Robot Manipulation via Terminal State Regularizationcited by 46
  17. 2023Hierarchical Planning and Control for Box Loco-Manipulationcited by 44
  18. 2023PMP: Learning to Physically Interact with Environments using Part-wise Motion Priorscited by 38
  19. 2023Neural Categorical Priors for Physics-Based Character Controlcited by 38
  20. 2022Learning to Use Chopsticks in Diverse Gripping Stylescited by 29
  21. 2023Composite Motion Learning with Task Controlcited by 28
  22. 2022ControlVAE: Model-Based Learning of Generative Controllers for Physics-Based Characterscited by 22
  23. 2024Strategy and Skill Learning for Physics-based Table Tennis Animationcited by 20
  24. 2021Discovering Diverse Athletic Jumping Strategiescited by 16
  25. 2023AdaptNet: Policy Adaptation for Physics-Based Character Controlcited by 16
  26. 2021A GAN-Like Approach for Physics-Based Imitation Learning and Interactive Character Controlcited by 13
  27. 2022A Survey on Reinforcement Learning Methods in Character Animationcited by 12
  28. 2023Synthesize Dexterous Nonprehensile Pregrasp for Ungraspable Objectscited by 10
  29. 2023An HCI-Centric Survey and Taxonomy of Human-Generative-AI Interactionscited by 9
  30. 2023Editing Motion Graphics Video via Motion Vectorization and Transformationcited by 8
  31. 2025PhysicsFC: Learning User-Controlled Skills for a Physics-Based Football Player Controllercited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 6.

6 papers.

  1. 2026Perceptive Humanoid Parkour: Chaining Dynamic Human Skills via Motion Matchingcited by 19
  2. 2026HumanX: Toward Agile and Generalizable Humanoid Interaction Skills from Human Videoscited by 14
  3. 2026Learning Human-Like Badminton Skills for Humanoid Robotscited by 3
  4. 2026MUSIC: Learning Muscle-Driven Dexterous Hand Controlcited by 1
  5. 2026MOCHI: Motion Enhancement of Collaborative Human-object Interactionscited by 1
  6. 2026ComplexMimic: Human-Scene Interaction Imitation in Complex 3D Environmentscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024SkillMimic: Learning Basketball Interaction Skills from Demonstrations
  2. 2025Catching Spinning Table Tennis Balls in Simulation with End-to-End Curriculum Reinforcement Learning
  3. 2025PACE: Physics Augmentation for Coordinated End-to-end Reinforcement Learning toward Versatile Humanoid Table Tennis
  4. 2023PlayBest: Professional Basketball Player Behavior Synthesis via Planning with Diffusion
  5. 2022Latent Plans for Task-Agnostic Offline Reinforcement Learning
  6. 2019Relay Policy Learning: Solving Long-Horizon Tasks via Imitation and Reinforcement Learning
  7. 2022GoalsEye: Learning High Speed Precision Table Tennis on a Physical Robot
  8. 2025HITTER: A HumanoId Table TEnnis Robot via Hierarchical Planning and Learning