Roboticist KevinO has been up to some really amazing stuff with his ROS Hexapod “Golem”. We’ve highlighted this robot before, but it just keeps getting better. The leap he’s made using mathematical odometry via Rviz in ROS is worth noting. Point cloud data (registered as a fake laser scan at robot height), IMU, and visual odometry from the Depth Sensor data are compiled and used by the mapping and navigation stack, then subsequently passed to a Kalman filter (a linear quadratic equation used to observe measurements over time). That’s a pretty comprehensive stack, giving the robot a good sense of its place in the world (as in its position in space, not in the annals of history).
Here’s what the point cloud scan looks like in rviz.
And here’s what an odometry plot looks like.
This project has been pretty impressive every step of the way. Golem has many abilities, including a smooth walking gait, body leveling via the IMU, text-to-speech ability, and now a much improved mode of navigation. This project was originally noted to be a ROS learning project, and I can’t wait to see what Kevin learns next!