My ongoing thesis topic is the study of Bottom-up Control. This loop of 36 Dynamixel Servos is just one particular case for me to work with. The goal of this particular case is to conduct Bottom-up Shape Matching, like a shape found to the right. This is done by treating each joint as an independent robotic agent and the agents work together to achieve the high level goal via a consensus algorithm.
This robotic system can be likened to a decentralized swarm. Where each each independent agent only is able of sensing information from its immediate surroundings. So a servo only know its own angle, the angle of its neighbors, and where its neighbors believe they fit into the solution. Every step in this recursive process takes these three inputs to determine a belief of where the servo fits into the overall solution.
This is a general description of the project and how it will function. Currently, this project is in the hardware interface stage. Control of the servos has gone well, but further integration into the Graphical User Interface (GUI) is taking more of the time.
Once the GUI is in a functional state, the MATLAB code infrastructure for the algorithm to control the robot will be written. Where as the code and mathematics governing these actions will be written parallel over the next month or so.
Once this stage of the Shape Matching is functioning, this project will proceed to integrate orientation and position matching in parallel to the shape matching. This under taking is enabled with a Vicon localization system, and a non-linear friction material for the system to scoot across.
Bottom-up orientation control will add a fourth variable that takes into account the orientation of the servo. While the Bottom-up control location control will add a fifth variable taking into account the XY-plane position. The system will move between locations thanks to the non-linear friction material and some form of Bottom-up motion to move across the plane. This could take many forms, such as snake-like locomotion, amoeba-like locomotion, or any such motions that will result in overall movement of the system.
Both of these aspects of the Bottom-up control case study are considered a more advance stage due to its integration of the Vicon system, the non-linear friction locomotion, and the further complication of the Bottom-up control as a whole.