Your project is a scientific study. The objective (and your grade) is not based on the fitness of your final solution but rather on the way you conduct the project.
This could include (but it is up to you to define your study):
- A statistical analysis of evolution parameters. The best parameters will be different for different fitness functions. The chosen parameters can also significantly impact the speed of convergence to a good solution.
- A comparison of different fitness functions and discussion about what factors may result in a certain morphology and controller vs another.
- A comparison between your evolved robot and a benchmark such as the hand designed starfish robot, etc.
- A comparison of evolution for multiple goals performed in one single step vs performed in multi-step evolution.
Remember:
- You should be writing your own fitness functions, e.g. explore the log files or use console.log() in the .js scenario file to see what sensor / position values you have available to include in your fitness function and what values they take under different conditions. Artificial evolution will only improve fitness for the function you write, a badly designed function can result in poor performance for the task no matter how long you evolve for.
- The solution that is found will almost certainly not be the global optima due to hardware and time constraints in addition to the size of the fitness landscape.
- You can do multi-step evolution for evolving a robot when the task is complex. How you group the tasks to evolve for at each step and in what order you perform each step evolution should be carefully considered as it can impact your final result.
