Their story, in their words…
What we did
We first created a simulation of the project so we could develop the maze solving algorithm. The robot in the simulation used Breadth-depth search algorithm and we then developed a q-learning algorithm.
Then began to build the maze solver robot. We used IR sensors to detect contrast allowing us to follow the maze path . For Goal detection (if the maze has been solved or not) we used an ultrasonic sensor. In terms of movement , we used 2 DC motors and a caster wheel.
Using arduino , we then coded the logic in the arduino language. We took multiple inputs from the sensors and based on what was inputted the motors would move at different speeds and directions.
For the physical robot, we started off with the left hand rule to solve the maze. So far that is the only algorithm we could transfer to the physical robot.
To learn machine learning and Reinforcement learning , I decided to take Andrew Ng’s specialisation course.For learning the robotics side of things , I mainly followed a youtube course and followed my inquisitions when met with obstacles.
How it worked out
Overall, the simulation and making of the robot was pretty quick.
Using more advanced algorithms than the left hand rule for the physical robot turned out to be harder than expected. Mainly because of the problems a robot has in the real world compared to a simulation which we did not take into account .
How we felt the project played out
I think it was good, especially applying code to something not digital .However with what we have learnt , I think we can improve the robot with a more advanced logic after mark evison. A possibility we did not consider during Mark Evison was using decision trees to solve the maze, which with the physical robot ,may turn out to be a more realistic and more optimal solution .
What we learned
We learned how to cooperate on a coding project, basics of machine learning which showed us the importance of maths in computer science.
It definitely showed me that artificial intelligence and robotics is something I want to do in the future. Currently , expanding on the maze solver and making more projects ,particularly projects that solve real world problems, is the main focus.
