Game Development and Neural Network
- FlapMind is a desktop game designed for Windows, developed in C++. Inspired by the classic Flappy Bird, this project reimagines the game with modern features such as multiple game modes and customizable themes.
- It not only revitalizes a classic game with new functionalities but also offers a practical opportunity to explore AI techniques, integrating NN and GA to demonstrate their application in gaming.
- As aspiring software developers, we are driven by the desire to expand our knowledge and skillset beyond the traditional boundaries of programming.
- The development of this project is fueled by the need to gain hands-on experience in integrating AI and machine learning techniques into a game-based environment.
- Game scenarios, with their controlled yet challenging nature, provide an ideal testbed for evaluating and validating AI algorithms, as they mimic real-world decision-making processes in a simulated setting.
-
In the game the player controls a bird that constantly falls due to gravity. To prevent the bird from crashing into the ground, the player must use the Spacebar or Up Arrow Key to make it flap and stay aloft. The goal is to navigate through gaps between pipes to achieve the highest score.
-
FlapMind utilizes Feed Forward Neural Network (FFNN), which is the simplest forms of neural network.
-
FFNNs operate with a straightforward architecture where data flows in a single direction—from input to output. This unidirectional flow of information simplifies the design and understanding of the network.
- Genetic Algorithms (GA) are optimization techniques inspired by the principles of natural selection and evolution. They work by simulating the process of natural evolution to solve complex problems.
- To enhance the performance of the Feed Forward Neural Network (FFNN) used in FlapMind, a Genetic Algorithm (GA) is employed.
- The GA helps in optimizing the neural network by evolving a population of solutions over time.
-
Crossover is performed by combining the genetic material (weights and biases) of the two best-performing individuals. This process involves exchanging portions of the parents' genetic code to create new offspring.
-
For FlapMind, a crossover rate of 49% is applied, meaning that there is a chance of 49% that the offspring will inherit traits from both parents. To introduce random variation and genetic diversity, each offspring has a 2% chance of undergoing mutation, where random adjustments are made to a portion of its weights.
