Locomotion & Controls Engineer

Innefu Labs; New Delhi; On-site; INR 900,000 - 1,500,000; Full time

Job summary

Innefu Labs invites a Locomotion Controls Engineer to advance legged locomotion research. You will design gait controllers in simulation, train them for mud, rock, snow, and stairs, and port them to real robots.

Required experience includes 3-8 years and advanced degrees in mechanical/electrical/robotics. Proficiency in C++, Python, and multiple physics simulators is essential, with hands-on hardware results.

Qualifications

• C++ and Python required.
• Rigid-body dynamics and control theory: MPC, whole-body control, or RL for locomotion (PPO).
• At least one physics simulator: Isaac Sim / Isaac Lab, MuJoCo, PyBullet, Gazebo or RaiSim.
• Has run a controller on real robot hardware, not only in simulation.

Responsibilities

• Build the simulation environment with randomised rough terrain.
• Develop gaits (walk, trot, stairs, crawl) using reinforcement learning, model predictive control (MPC), or both.
• Transfer controllers from simulation to the real robot (sim-to-real) and tune them in field tests.
• Write state estimation (body pose and velocity) from IMU and joint sensors.
• Specify actuator and leg requirements to choose the off-the-shelf supplier.

Skills

C++

Python

MPC

Whole-body control

Reinforcement learning

Education

MTech/PhD with a thesis on legged locomotion

BTech/MTech/PhD in Mechanical, Electrical, Aerospace, Robotics or Control Systems

Tools

Isaac Sim

MuJoCo

PyBullet

Gazebo

RaiSim

Job description

Job Summary

Locomotion Controls Engineer

Role Overview: Makes the robot walk, climb and recover on mud, rock, snow and stairs designing and training the gait controller in simulation and transferring it to the real robot.

Experience

3-8 years

Education

MTech/PhD with a thesis on legged locomotion. BTech/MTech/PhD in Mechanical, Electrical, Aerospace, Robotics or Control Systems. Very few people in India have built legged robots, so a strong humanoid, exoskeleton or legged-research background also counts.

Responsibilities

• Build the simulation environment with randomised rough terrain.
• Develop gaits (walk, trot, stairs, crawl) using reinforcement learning, model predictive control (MPC), or both.
• Transfer controllers from simulation to the real robot (sim-to-real) and tune them in field tests.
• Write state estimation (body pose and velocity) from IMU and joint sensors.
• Specify actuator and leg requirements to choose the off-the-shelf supplier.

Must have

• C++ and Python.
• Rigid-body dynamics and control theory: MPC, whole-body control, or RL for locomotion (PPO).
• At least one physics simulator: Isaac Sim / Isaac Lab, MuJoCo, PyBullet, Gazebo or RaiSim.
• Has run a controller on real robot hardware, not only in simulation.

Good to have

• Hands-on with commercial quadrupeds (Unitree, ANYmal, Spot, Deep Robotics) or their SDKs.
• Open-source legged stacks: legged_gym, OCS2, Pinocchio, Drake, Crocoddyl.
• Publications at ICRA, IROS, CoRL, RSS or Humanoids.
• Kalman filters / state estimation; terrain-aware foothold planning.
• PyTorch, GPU training, domain randomisation.

Job Details

• Job Type: Full time
• Location: Delhi, India

Disclaimer: This job posting and Location has been aggregated from external source. Role details, content, and availability are subject to change. Applicants are advised to confirm the latest information directly on the company website before applying.