Learning Options for Compositional Motor Control with Adapter Banks

We ask whether the representational hierarchy seen in perception, from local primitives such as edges to higher level structures such as parts and objects, can be established for motion. In humanoid control, low level actions specify immediate motor commands, while meaningful behavior is organized over longer temporal scales, including contacts, gait fragments, balance recovery, reaching, and whole body skills. We introduce MotionPyramid, a hierarchical action representation that learns such structure from motion data. Starting from a motion tracking teacher, it trains a recursive stack of latent decoders: low level latents decode to immediate full body motor commands, while higher level latents unfold through lower levels into temporally extended motion programs. After pretraining, the hierarchy is frozen and reused by downstream reinforcement learning policies as a family of action interfaces at different control resolutions. Experiments show the learned levels form a motion hierarchy: coarser interfaces improve early learning and motion regularity by constraining exploration to structured segments, while finer interfaces preserve feedback control and final task precision. Representation probes show the hierarchy supports traversal, interpolation, transition, and qualitative composition, exposing editable control handles across temporal scales. Finally, we introduce Residual Interfaces, letting a downstream policy maintain coarse, segment level, and frame level residual commands through the frozen hierarchy. Analogous to residual or skip connections in deep networks, this allows coarse motion programs and fine residual corrections to coexist within one controller. MotionPyramid shows that motion, like perception, can be organized into a reusable multi level representation, providing structured abstraction without sacrificing controllability.

Generalized morphology control requires a single policy to transform information across limbs with different physical roles, coordinate whole-body motion, and remain efficient as body size grows. Existing communication mechanisms address these requirements only partially. We introduce RecMorph, a topology-guided spatial recurrent architecture that uses recurrent sequence computation to jointly perform cross-limb communication and representation transformation. A depth-first traversal converts the kinematic tree into a morphology-derived sequence, along which shared bidirectional transitions progressively transform limb information before action decoding. Residual preservation, RMS normalization, and input-dependent channel modulation stabilize this repeated spatial transformation, yielding linear token complexity at fixed model width and depth. Across five UNIMAL tasks, RecMorph achieves the strongest mean final training performance among the evaluated generalized morphology controllers and the highest measured inference throughput on FT, while generalizing to unseen variations and bodies with up to 30 limbs. We further migrate representative generalized controllers from UNIMAL benchmarks to a four-platform quadruped setting. RecMorph achieves the best macro-averaged performance under nominal and high friction, reduces nominal velocity RMSE by 43.5% relative to specialist MLPs, and one shared policy completes 40 physical Go1/Go2 trials without falls. These results show that topology-guided recurrent transformation provides an effective and efficient communication mechanism for Generalized Morphology Control and remains effective when transferred from procedural bodies to physical robot platforms. Code and experimental resources are publicly available at https://github.com/quanruirao/RecMorph.

Pretrained imitation policies have become a strong foundation for robot manipulation, but they often require online improvement to overcome execution errors, limited dataset coverage, and deployment mismatch. A central question is therefore how reinforcement learning (RL) should adapt policies after offline pretraining. Existing lightweight methods commonly apply residual corrections directly in action space, but this often leads to noisy and poorly structured exploration. In this work, we propose Z-Perturbation Reinforcement Learning (ZPRL), an approach that steers pretrained policies through a compact bottleneck latent rather than through policy weights or output actions. During offline training, we augment the policy with a plug-and-play variational information bottleneck (VIB) module to extract a task-relevant latent interface from observation embeddings. During online finetuning, the base policy is frozen and RL learns only a residual perturbation on this latent, whose decoded representation conditions the frozen action generator. We instantiate ZPRL on flow-matching policies and evaluate it on eight simulation tasks and four real-world tasks. Across diverse manipulation settings, ZPRL improves both sample efficiency and final performance over strong post-training baselines. In the real world, ZPRL improves the average success rate on four tasks by 33.7% over imitation base policies while producing smoother exploration behaviors than an action residual counterpart. These results suggest that a compact, task-aligned bottleneck latent provides an effective interface for online RL adaptation. More videos can be found at https://manutdmoon.github.io/ZPRL/.