Lorcan the Quadruped Robot

thingiverse

Lorcan is a small, agile robot designed to be low-cost (< $1k USD) and easy to produce. This compact quadrupedal bot draws inspiration from a well-known robotic design but features a unique twist - a simple gear system for direct drive operation. The brain of Lorcan consists of four ODrive motor controllers and eight encoders, which work together to enable the robot's movements. The ultimate goal is to utilize machine learning algorithms to teach Lorcan how to walk instead of relying on complex control systems. To achieve this, a bottom-facing camera with optical flow capabilities will generate a reward signal for reinforcement learning. A Raspberry Pi 4 powers the system, providing the necessary computational resources for efficient processing. Updates to the design have been made to address minor issues, including the addition of four spacers that can be printed separately or incorporated into the femurs. The project has progressed significantly, with the robot's mechanics now complete and wiring/electronics remaining. A new progress shot showcases the current state of the bot, with control over the limbs possible but basic walking capabilities still under development. Once these gates are established, OgmaNeo will be employed for reinforcement learning after a Pi Camera is mounted.