Advanced Robotics
Ant Colony Swarm
As the final class in the robotics series, the requirements for this project were slightly more rigorous than the past two classes. While sticking to the theme of choosing the path of the project, some more restrictions were added. This project required major adaptions to previous robots or build a full robotic system from scratch while also incorporating novel ideas rarely, or never, seen before with the cause of solving a particular problem. For my project, along with my teammate, this was interpreted as the creation of our own swarm of microrobots. This was built upon the search and locate problem seen in several industries such as search and rescue, fire management, criminal justice, ecology, energy and so many more. In order to ensure novelty, it was decided to develop a swarm algorithm from scratch rather than utilize a predesigned algorithm.
For the swarm algorithm, the search functions were largely based on previous ant colony algorithms with some portions of beehive algorithms mixed in for added robustness. This allowed each robot to function independently of the colony to locate the goal and then perform a gathering action upon reaching said goal. All of this was managed by leaving digital "pheromones" that could be detected by the other robots and traced to goal points as in a real ant colony. The random wandering of the robot was controlled through a Control Theory based algorithm that managed orientation and position while transitioning between randomly chosen points within the designated space.
After a few iterations the chassis was largely minimized in design to make the footprint smaller as well as shortening print time on the 3D printers being utilized. For the brains of the robot, an ESP8266 WI-FI enabled microcontroller was chosen to allow for WI-FI based communication between the robots and a central computer used to store bulk information needed by the robots such as maps and environmental obstacles that had been previously located.
Overall, the project did much of what it set out to do and exceeded expectations of both the team and the professor in what was accomplished. WI-FI interfacing worked to transmit data back to the home base station and was used as a diagnostic tool but was unable to be used as a communication tool with other robots. The pheromone trails also appeared to work as hoped but a communication error left this an unresolved success. The best functionality of the robots was the orientation and position based random wandering navigation. This worked better than expected with consideration of the high error percentages caused by the tolerances in the odometry being tracked from wheel axle.