EvoBots: Artificial Life Evolution Simulation
EvoBots is an artificial life evolution simulation using neural networks and genetic algorithms to evolve bots with artificial intelligence in a virtual world.
Artificial Intelligence
To survive and thrive in the EvoBots simulation, bots have to evolve a basic form of artificial intelligence. All bots have a neural network as their brain and only the bots that figure out how to get energy from eating plants or attacking other bots will survive and ultimately reproduce to ensure the survival of their species. Complex behaviors emerge over time and have not been scripted. The goal of this project is to create basic artificial life in the process.
Neural Networks
The brain of each bot is a simple neural network with one hidden layer. Through inputs from the environment, the neural network decides which action to take in any given situation. This architecture is enough to create simple functional artificial life forms in a virtual world. The simulation can be run with a pre-trained neural network to demonstrate the proof of concept or evolving the bots from scratch based on random neural networks.
Evolution
Evolution and genetic algorithms are a central part of the EvoBots artificial life simulation. All bots are subject to natural selection and survival of the fittest. When bots create a copy of themselves through asexual reproduction, their neural networks and body attributes mutate slightly. These mutations can have a positive effect and allow them and their offsprings to be more successful in the simulation. Over time this process leads to well adapted bots.
Procedural Generation
Wherever possible EvoBots uses procedural generation to make the simulation more interesting. All bot bodies are generated based on simple rules and procedurally adding and removing body parts based on the bots DNA. This makes the bots look unique since the number of possible configurations is relatively high. The ecosystem and terrain are also procedurally generated which makes each simulation run a new experience.
Send Feedback
EvoBots is an evolving project and feedback is very much appreciated. Get in touch via social media or write an email at mail@evobots.ai