Adaptation of Neural-Fuzzy controlller applied to a LEGO MINDSTORM robot
The objective of this research is to utilize Fuzzy Controls to replace an existing controller used to balance an unmanned vehicle along the vertical axis. The outcome of the substitution using Fuzzy and Neural-Fuzzy controls was the successful design such that the gain controller exceeded the performance of the static gain on the LQR controller and the combination of the gain and controller mimicked the LQR. The application of both Fuzzy and Neural-Fuzzy controllers in this research is to highlight the strengths of both ease of application and their intuitive design. These controllers allow a human operator to focus on the simple tasks of driving the platform. This task in combination with the unmanned vehicle performing the complex task of balancing makes for a usable system that otherwise would not be possible.
The potential for a human operator performing this task of balancing a two-wheeled robot based on visual cues or the platform telemetry, would be difficult if not impossible. The use of an operator control station allows for a faster design process while providing a point at which the addition of an off-platform Ethernet controller can be added for distributed control as well as allowing the operator to tell the platform to take over the balance control locally on the platform. Simulations of the vehicle and controller were designed using MATLAB and Simulink; the application for the controllers is capable of being uploaded to the real-time unmanned vehicle for proof of concept.