Lucen offers a modular robotics platform that converts between a bipedal robot and a 7-DoF arm using a six-part swap, enabling both locomotion and manipulation research from a single kit. The system includes an NVIDIA GPU, open-source firmware under an MIT license, and supports standard training pipelines like MuJoCo, Isaac Sim, and LeRobot. Priced from $3,999, the kit targets researchers, students, and hobbyists working on reinforcement learning and control.
Funding
Funding not disclosed
Founders
Product
Problem
Researchers, students, and hobbyists in robotics often need separate, specialized hardware for locomotion and manipulation experiments, which is costly and space-intensive. Existing robot kits are typically fixed-purpose, limiting flexibility and forcing users to purchase multiple platforms to explore different research areas.
Solution
Lucen provides a fully modular robotics platform that transforms between a bipedal robot and a dexterous arm using the same actuators, controllers, and firmware. Users start with a bipedal configuration for training locomotion policies, then swap six parts with a hex key to convert the kit into a 7-DoF manipulator with a compliant gripper for manipulation research. The platform runs standard simulation and sim2real pipelines, including MuJoCo, Isaac Sim, and LeRobot, and includes an onboard NVIDIA GPU for compute-intensive tasks. Open firmware under an MIT license and public design files ensure full customizability and reproducibility.
Target Audience
Primary users are robotics researchers, university students, and hobbyists focused on reinforcement learning, control theory, and manipulation research who need a versatile, cost-effective hardware platform.
Features
- Modular design with reusable actuators, controllers, and firmware across bipedal and arm configurations
- Six-part swap mechanism enabling conversion between biped and 7-DoF arm in minutes with a hex key
- Onboard NVIDIA Super 8GB GPU with 128 GB SSD and WiFi module for edge compute
- 12 serial-bus torque-controlled actuators supporting single-leg mode for control theory education
- Compatibility with MuJoCo, Isaac Sim, sim2real, LeRobot, ACT, and SmolVLA training pipelines
- Open firmware under MIT license with public GitHub access to teleop tools and training notebooks