Deep Robotics

Founded by a team of Zhejiang University PhDs, Deep Robotics earned its reputation in the field rather than the demo hall: it was the first company in China to run fully autonomous substation inspections with quadruped robots. That inspection heritage shapes the whole catalog - these are working platforms with hot-swappable batteries, self-recovery from falls, and an open SDK, not show robots.

The Lite3 research line is deliberately modular. Every tier shares the same 12-degree-of-freedom chassis, 2.5 m/s top speed, and C++/Python SDK with ROS 1 and ROS 2 support - what changes is the stack on top. The Basic gives you the bare locomotion platform; the Basic AI unlocks reinforcement-learning gaits out of the box; the Venture adds expansion interfaces and power outputs for your own sensors; the Pro adds an NVIDIA Jetson and depth camera for perception work; and the Lite3 LiDAR adds 3D LiDAR with a pre-configured SLAM and navigation stack. You buy the capability you need and code against the same APIs at every level, with URDF models published for Gazebo, MuJoCo, and Isaac Sim.

The Lite3 series spans education and research budgets from the entry Basic to the perception-equipped Pro. All climb stairs, carry payloads on a standardized mounting deck, and run 1.5 to 2 hours per hot-swappable battery.

At the top of the range, the Lite3 LiDAR handles autonomous navigation research, and the Lynx M20 Pro is a different animal entirely: a wheel-legged industrial platform with IP66 sealing, a 15 kg working payload, dual 96-line LiDARs, and an operating range from -20 to +55 degrees C for outdoor inspection work. A spare Lite3 controller is also available.

Not sure which tier fits your lab or site? Our Deep Robotics quadruped comparison lays out payload, sensors, compute, and runtime for every model in one table.