This project was created for fun and studying purposes. These were the mains goals behind it:
-
Use the Reactive Message Orientation Object Orientation. A paradigm I conceived to try to reflect Reactive System design in a OO level to get the hang of it and get used to the difficulties that may occur. To do this, I elaborated two ground rules:
- Message oriented object communication: no public methods. Every communication has to be through a message mechanism (Event emitters) and serializable objects (dto, vo, and others). Objects are not aware of each other.
- Stateless messages: the messages should include every information it needs, so the receivers process them accordingly.
-
Study Neural Network. Find a mathematical function that receives inputs (bird y position, the closest pipe gap x position, the closest pipe gap y position) and results in a single output (should jump?).
-
Study Genetic Algorithms. Select the best citizens of every generation and create a new generation based exclusively on this and operations such as crossover and mutations.
-
Study Reinforcement Learning (q-learning). An agent (i.e. the bird) takes in a state and a reward from the environment, and based on these variables, the agent chooses the optimal action (flap/not to flap).
-
Study Simulated annealing. Wikipedia. Specifically, it is a metaheuristic to approximate global optimization in a large search space for an optimization problem.
-
Have fun. Yay.
If you want to, you can watch it live or/and defy the AI. I should warn you, it's not a good idea.
Remember to control the BLUE one (press SPACE BAR to jump and ESC to reinitialize round). Feel free to mess around with the url query params (Defaults to: mutationRate=0.01&populationPerGeneration=1500&relativeSelectedPopulationPerGeneration=0.01&qBirdsNumber=100)
