Evolutionary Reinforcement Learning in an Ecosystem Based Environment

Left is the neural network (http://alexlenail.me/NN-SVG/index.html) and right is an example of the observations from an agent,

Results

The training rewards (left) and evaluation rewards(right) for various methods. ERL and Q-learning perform similarly, while the hard coded agents performed the best.
The number of agents of each species as a function of time. The predators die out relatively soon.
Left is the real world predator prey models, and on the right is our model. We see the same oscillatory structure and out of phase cycles. The leftmost image was found here, licensed under the CC Attribution-Share Alike License 3.0.
Carrying capacity — The environment can only support a certain number of agents due to the food growth rate.
Competition between different types of predators, and the random predators die out because the ERL predators out-compete them.

Limitations

Conclusion

References / Acknowledgements

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store