Advanced Hexapod Robot: autonomous walking with sensor fusion

What the experiment does

A physical six-legged robot we built and programmed. It tests how a machine works out where it is from noisy sensor readings and how it plans its way to a destination. Each leg has its own motors, so the software computes every leg's movement in real time. Warehouse robots, self-driving vehicles and other autonomous machines need both skills.

How it works

The robot receives a destination. It reads its sensors continuously, and every reading carries some noise. A mathematical filter fuses those readings into one estimate of the robot's position and heading that the planner can rely on. The navigation planner plots the route to the destination from that estimate. Each step along the route becomes a gait pattern that decides which legs lift and which stay on the ground. The software computes the joint angles of every leg in real time, so the motors move together in the planned direction. After each step new readings arrive, and the robot corrects its estimate and carries on.

What we can build for you

• We fuse several sensors into one estimate your system can rely on, for any device that tracks position or motion.
• We build the navigation layer that plans a device's route to its destination and replans whenever its position changes.
• We write real-time motor control that runs on a small, low-cost computer inside the device itself.

How it is built

The control loops run on a Raspberry Pi mounted on the robot, so the robot carries its own computer. ROS 2 links the parts of the software, so readings and commands pass between them in real time. A Kalman filter fuses the noisy readings into one estimate of position and heading. Nav2, the standard navigation stack for ROS 2, plans the route from that estimate.

• ROS 2
• Raspberry Pi
• Kalman Filter
• Nav2