Even though we are quite pleased with the performance of Marvin there are some issues that we would like to improve in the future if the opportunity presents itself.

Marvin is able to balance, drive a around randomly and avoid obstacles. He is also able to be remote controlled via Bluetooth. However these two behaviours do not operate well together due to resource sharing problems in the threads. We did not find a solution to this problem, but keeping them both at the same time had a high priority in the group since it is an essential part of the behaviour architecture. The problem might be solved by changing to another framework with a different thread implementation, but we have not investigated this further.

We chose to use a PID controller because it had been introduced during the course, but it could be interesting to investigate some other controllers and see if it was possible to do some analytical work to see which one is best. The controllers of interest could be LQR controller or Pole-Placements. These are mere a few amongst many. We are curios as to whether it could be possible to further optimize the PID control parameters for better balancing robustness, or if it has been pushed to its limits whereas we would have to turn to other controller algorithms to achieve better results.

In order to try and further optimize the PID control parameters (or any other parameter of interest), it would be nice to have an online tuning method, through a PC GUI application, communicating with the robot by means of bluetooth.

In lab 11¹⁾ we outlined the project with its path of progression, problems and goals. To summarize it we wanted to “be able to show a bluetooth remote controlled gyro-balancing robot, driving autonomously when not controlled, capable of avoiding obstacles.”.
During the entirety if the unwinding of the project, we kept the original goals in mind, and tried to keep as close to them as possible. We did not expect, but hoped, to accomplish all of them, however contrary to our expectations, we managed to achieve all, but one (the GUI).

Conclusively we consider this project as successful.