人形机器人运控工程师/研究员 / Humanoid Control Engineer/Researcher

岗位使命

构建人形机器人的“运动小脑”：研发类似 SONIC 的通用全身运动模型，将任务目标、机器人状态和运动指令转化为连续、协调、可在真机执行的全身动作。

你将负责

  • 研发类似 SONIC 的通用全身运动模型，基于强化学习学习可复用、可组合、可迁移的全身运动技能。
  • 研究 Motion Tracking、运动技能表示、动作生成、运动重定向、状态估计和 Sim-to-real。
  • 探索运动小脑与 VLA、高层任务规划及机器人实时执行系统之间的接口。
  • 参与真实人形机器人的遥操作系统和遥操数据采集，建立高质量的行为数据闭环。
  • 完成从仿真、强化学习策略训练到真实硬件部署的完整闭环，处理接触、摩擦、时延、噪声、模型误差和硬件约束等真实问题。

我们希望你

  • 熟悉强化学习、Whole-body Control、Motion Tracking、Motion Retargeting、运动技能学习或 Sim-to-real 中的一种或多种。
  • 有使用强化学习训练人形机器人全身运动策略的经验，理解奖励设计、Domain Randomization 和真机迁移。
  • 熟悉机器人运动学、动力学和状态估计，具备扎实的 C++、Python 和 Linux 工程能力。
  • 有真实机器人调试经验，能够独立定位软硬件链路中的问题；有遥操作或遥操数据行为学习经验者优先。

Humanoid Control Engineer/Researcher

Mission

Build the humanoid's "motor cerebellum": develop a SONIC-like general whole-body motion model that turns task goals, robot states, and motion commands into continuous, coordinated, whole-body actions executable on real hardware.

What You'll Do

  • Develop a SONIC-like general whole-body motion model, learning reusable, composable, and transferable whole-body skills via reinforcement learning.
  • Research motion tracking, motor skill representation, motion generation, motion retargeting, state estimation, and sim-to-real.
  • Explore the interfaces between the motor cerebellum and VLA models, high-level task planning, and real-time execution systems.
  • Contribute to teleoperation systems and teleop data collection on real humanoids, building a high-quality behavioral data loop.
  • Own the full loop from simulation and RL policy training to real-hardware deployment, handling contact, friction, latency, noise, model error, and hardware constraints.

What We're Looking For

  • Familiar with one or more of: reinforcement learning, whole-body control, motion tracking, motion retargeting, motor skill learning, or sim-to-real.
  • Experience training humanoid whole-body motion policies with RL; understanding of reward design, domain randomization, and real-robot transfer.
  • Solid grasp of robot kinematics, dynamics, and state estimation, with strong C++, Python, and Linux engineering skills.
  • Hands-on debugging experience on real robots, able to independently diagnose issues across the hardware-software stack; teleoperation or teleop-data behavior learning experience is a plus.