ImpedanceMimic:用学习到的阻抗作为动作空间,实现力精确的运动模仿
ImpedanceMimic: Learned Impedance as Action Space for Force-Accurate Motion Imitation
一篇把关节刚度和阻尼变成强化学习动作空间的研究,解决了固定增益 PD 在模仿动作时对不上接触力的老问题。129 人、六种活动的数据验证,还能零样本迁移到人形机器人。做仿真、人形控制或动捕动画的值得细看。
ImpedanceMimic 是一项机器人与生物力学方向的新研究,核心发现是传统基于物理的运动模仿依赖固定增益 PD 控制,而固定增益会在运动跟踪与力跟踪之间造成折衷,因为接触与摆动阶段对关节刚度和阻尼的要求相互冲突。研究者提出一个强化学习策略,从运动和力测量中调制每个关节的刚度、阻尼和位置设定点,推理时只需要运动数据。该方法将增益表示为惯性缩放先验的倍数,使单一策略可以控制不同个体的生物力学模型,无需显式人体测量输入。在 129 名参与者、六种活动的评估中,ImpedanceMimic 在运动学和力保真度上都超过了匹配的固定增益 PD、直接力矩控制、肌肉驱动控制等已有模仿控制器,接近专门的运动到力回归器。学习到的阻抗还形成了跨身体的低维运动状态表示,能零样本迁移到未见过的人类和一台人形机器人,且表现出刚度随负载增大、阻尼预判卸载的泛化规律。
对做物理仿真、人形机器人控制或动作捕捉驱动动画的开发者来说,这项工作给出了一个新思路:把关节阻抗本身当作策略的输出空间,而不是只模仿轨迹。相比固定增益 PD 控制在力精确任务上的固有局限,这种方式在接触丰富的场景(如搬运、跌倒防护、人机交互)中更容易复现真实的接触力,代价是需要力测量数据做训练。
ImpedanceMimic: Learned Impedance as Action Space for Force-Accurate Motion Imitation
Recovering the forces behind a motion is ill-posed: different controllers can reproduce the same movement while generating different contact forces. Physics-based motion imitation resolves this ambiguity through simulation, but relies on fixed-gain proportional-derivative (PD) control. We show that fixed gains impose a trade-off between motion and force tracking, as contact and swing place conflicting demands on joint stiffness and damping. We introduce ImpedanceMimic, a reinforcement-learning policy that modulates per-joint stiffness, damping, and position setpoints from motion and force measurements, but requires only motion at inference. Across 129 participants and six activities, ImpedanceMimic outperforms matched fixed-gain PD under every integration scheme, direct-torque, muscle-actuated and existing imitation controllers in kinematic and force fidelity, approaching a dedicated motion-to-force regressor. By expressing gains as multiples of inertia-scaled priors, a single policy controls subject-specific biomechanical models without explicit anthropometric input and transfers zero-shot to unseen humans and a humanoid robot. The learned impedance forms a low-dimensional representation of motor state, shared across bodies, that generalizes to withheld activities: stiffness scales with load, and damping anticipates unloading.
This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include ImpedanceMimic, reinforcement learning, humanoid robots, biomechanics.
Source attribution: curated from arXiv cs.LG (@Matthias Jammot, Pembe Gizem Özdil).