Grammatical Evolution based Controller Design for Vehicle Robot Action Learning

Compared to humans that naturally learned to accomplish a task by learning from psychomotor skills, robots perform a complete task by dividing tasks into smaller executable tasks and run in sequence. Creating a control program for robots needs explicit definition of each action to handle inputs from sensors and produce outputs to actuator. This project aims to analyze Grammatical Evolution(GE), a naturally inspired evolutionary algorithm derived from Genetic Programming (GP) to automate the process of producing controller instructions for Vehicle Robot. The advantage of this algorithm is the ability to separate searching space and generated program, thus eliminating human criteria bias when creating a robot controller while able to find more optimized and complex skills learning. The evaluation is done through partially observed maze problem which objective to find goal without defining starting orientation and discoverable path along the way. Several Evolutionary parameters are compared to find the optimized value that suits the application of robot. Then, several paths generated from the program are compared and discussed to determine the efficacy of the algorithm in locating the optimal program. The proposed method is more pragmatic as it focusses on the actual actions sequences on how the program is developed automatically to solve the problem.