An open source framework for building new quadrupedal robots, CHAMP is based on a hierarchical controller design for highly dynamic locomotion first implemented on the MIT Cheetah robot. The framework, also developed for implementing new control algorithms, includes a setup assistant to configure newly-built robots and features full autonomy using the ROS navigation stack.
While hardware documentation is not currently available for demo bot construction, the software stack is, and you don’t need a physical robot to run the demos — you can run everything in an available simulation environment. Additionally, the documentation includes a collection of pre-configured Unified Robot Description Formats (URDFs), including Anymal, MIT Mini Cheetah, Boston Dynamic’s LittleDog, and SpotMicroAi.
The algorithm that forms the basis for CHAMP was developed at MIT for high-speed running in a quadrupedal robot design. This involves a few challenges, namely locomotion stability, control of ground reaction force, and coordination of all four limbs. The strategies used in the control algorithm are as follows: leg impedance control provides programmable virtual compliance of each leg; the forces exerted on the ground by each of the four legs was determined and controlled using the equilibrium hypothesis. The gait pattern modulator imposes the desired footfall sequence to attain limb coordination. This algorithm was able to demonstrate high-speed trot running on a treadmill during experiments conducted with the MIT Cheetah robot.
Like its predecessors, CHAMP can walk autonomously using automatic gait optimization and perform tricks. The Mini Cheetah was the first-ever quadrupedal robot to do a backflip, and this open source framework serves as a jumping-off point to build on that locomotive ability. The setup assistant available via GitHub generates a robot configuration package, with instructions for configuring own. The generated package includes the microcontroller header files for gait and lightweight robot description, for robot builds that use microcontroller to run the quadruped controller, as well as URDF path, joint and links maps, gait parameters, hardware drivers, and navigation parameters. The collection of pre-configured bots can serve as a reference, and are also available to download and try.
Once hardware documentation is available, it should detail a demo bot using accessible components, so it's simple to build and test at home. In the meantime, the Gazebo demo environment is a useful sandbox for a quick-start to a summer robotics project.