Project - Evolution of agents

This project focuses on the evolution of agents using the Evolution Gym suite. To get started with evogym, see the documentation.

Neuroevolution in evogym notebook

Notebook for gifs

You will need to evolve movement policies for three tasks independently:

Walker-v0 (easy)
Thrower-v0 (medium)
Climb-v2 (hard)

You have a budget of 10000 evaluations for evolution (for example, a population of 10 for 1000 generations). Each evaluation can have 500 maximum steps, but you are encouraged to reduce this and the total number of evaluations while making algorithm decisions to have faster results. You choose the evolutionary algorithm, gene representation, and evolutionary hyperparameters, but you must demonstrate that you only used the allocated evaluation budget. The goal is to obtain the best score independently on each task. Scores should be shown in your final presentation as the average best score over at least 2 independent evolutions.

For each task, points will be allocated to teams in the following manner:

1st place: 5 points
2nd place: 4 points
3rd place: 3 points
4th place: 2 points
5th place: 1 point

Project presentations will take place on Tuesday, May 7th.

You can use the code provided during class for your evolutionary algorithms, and you can also use any code online. Some popular libraries are: