Ekumen connects a MuJoCo-trained manipulation policy to physical hardware
Ekumen described a client project that moved a reinforcement-learning manipulation policy from MuJoCo into a reproducible ROS 2 hardware stack.
Evidence notes
- The ROS 2 system connected the policy to a robot arm, gripper, motor controllers, perception components and operator tools.
- Ekumen built a reproducible development environment and shared-control interfaces for setup and demonstrations.
- The project addressed the transition from simulation-trained policy execution to integrated physical hardware.
Company context
Ekumen is a robotics and advanced-software engineering company that helps organisations develop, simulate and deploy autonomous systems. Its teams work across ROS, navigation, perception, simulation, web interfaces and cloud infrastructure for mobile robots, industrial systems and autonomous vehicles. The company also contributes open-source tooling and provides embedded engineering for product teams.