This is a ROS based controller for a quadrupedal robot intended to run on a Raspberry Pi. There is a gait generator, monocular visual odometry, sparse SLAM (in progress), and a simulator for the robot itself.
The robot will eventually have integrated elements of machine learning systems. The gait and path planning will be influenced via different ML algorithms.
I used ubuntu Mate which can be installed directly on to a raspberry pi. From there install ROS kinetic using 'sudo apt-get install ros-kinetic-desktop-full'
This downloads ROS and all dependencies necessary for this project.
The simulator can be run by calling 'roslaunch quadquad_gazebo basicworld.launch'
The gait controller can be run by calling 'python /path/to/gait_controller.py'
The vision odometry (egomotion) and SLAM can be run by calling 'rosrun quadquad_vision quadquad_vision_node'
The goal of this project is to integrate machine learning with conventional robotics. The quadruped platform seemed to be a very interesting way to accomplish this and allows a steady trend of integration.
The robot does not currently have a depth sensor, but a sharp infrared depth sensor will be added soon to better create an occupancy grid for navigation.