Dots Simulation using Genetic Algorithm - Part 2

This tutorial enhances a genetic algorithm simulation by adding features like elite highlighting, increased obstacle complexity, a "Reached" counter, and crossover breeding. Let's break down the improvements.

Part 1: Visual Enhancements and Obstacle Complexity

The simulation is upgraded to visually distinguish elite dots (those performing best in the previous generation) by coloring them blue. This is achieved by adding an is_elite boolean parameter to the Dot class's draw method and conditionally applying the blue color. The Population class's draw method is modified to pass this boolean based on whether a dot is in the elites list.

Obstacle generation is refactored for greater flexibility. Obstacle and Goal classes are moved to a separate obstacles.py file, promoting cleaner code organization. A constants.py file is introduced to hold global variables like screen dimensions and population size, preventing redundancy across files. Multiple obstacle configurations (OBSTACLES0, OBSTACLES1, OBSTACLES2, OBSTACLES3, OBSTACLES4, OBSTACLES5) are defined in obstacles.py, allowing easy switching between different challenge levels. The main script imports these configurations and selects the desired one. A check is added to ensure the goal is always present, even when using obstacle lists generated via list comprehensions (like OBSTACLES4).

A "Reached" counter is added to display the number of dots that successfully reached the goal in the previous generation. This is implemented by modifying the generate_next_generation method in the Population class to count and return this value. The main loop then displays this count on the screen.

Simulation with elites highlighted and obstacles

Simulation running OBSTACLES0 obstacles

Part 2: Implementing Single-Point Crossover

The simulation transitions from replication to single-point crossover for offspring generation. A crossover class method is added to the Dot class. This method takes two parent dots as input, selects a random crossover point, and creates two offspring by combining portions of each parent's movement sequence (represented as a list of direction vectors). The generate_next_generation method is updated to utilize this crossover method, generating pairs of offspring instead of single clones. Mutation continues to be applied to the offspring.

Single-point crossover

The improved simulation offers enhanced visualization, adjustable difficulty, and a more sophisticated breeding mechanism, making it a more robust and insightful example of a genetic algorithm. Future improvements mentioned include save/load functionality and speed optimization. The author also encourages joining their Discord community for further collaboration.

The above is the detailed content of Dots Simulation using Genetic Algorithm - Part 2. For more information, please follow other related articles on the PHP Chinese website!