WallerLocal implements genetic algorithm-based evolution of robot behavior for tracking and following a moving ball in an environment with obstacles. The robot must develop efficient ball-following strategies using only local perception.
The robot navigates a matrix environment to reach a ball while:
- 200x200 grid world
- Avoiding multiple fixed obstacles
- Continuously tracking the moving ball with physics-based behavior
- Minimizing movement continuous movement and repositioning
- Adapting to random initial positions
- Local vision (30-degree cone)
- 5-degree angular turns
- Forward/backward movement
- Ball detection and tracking
- Obstacle avoidance
#define GENS 51 // Number of generations
#define POPULATION 500 // Population size
#define CROSSING 350 // 70% of population for crossover
#define REPRODUCTION 150 // 30% of population for reproduction
#define HEIGHT 200 // Matrix height
#define WIDTH 200 // Matrix width
#define RUNS 1 // Number of tests per individual
#define EXECUTE 2000 // Number of tree executions per test
#define LIMIT 1000 // Individual complexity limit
#define ANGLE 5 // Robot turning angle
#define HIT_DISTANCE 1 // Ball touch detection distance
#define VIEW_ANGLE 30 // Local vision angleThe fitness function considers:
- Initial distance between robot and ball
- Distance after each ball touch
- Number of robot steps (movement efficiency)
- Number of successful ball hits
PROGN3 (3): Executes three branches in sequencePROGN2 (2): Executes two branches in sequenceIFWALL (I): Executes left branch if wall detected, right branch otherwiseIFBALL (C): Executes left branch if ball in sight, right branch otherwise
WALKFRONT (F): Move forwardWALKBACK (B): Move backwardRIGHT (R): Turn right (depends on ANGLE)LEFT (L): Turn left (depends on ANGLE)ALIGN (A): Orient towards ball (max 30 degrees)
- Number of generations: Set by GENS
- Population size: Set by POPULATION
- Reproduction probability: Set by REPRODUCTION
- Crossover probability: Set by CROSSING
- Mutation probability: 0%
- Individual complexity is limited by LIMIT parameter
- Uses local vision system
- Includes ball tracking function
- Features memory-efficient design
- Implements obstacle detection
This project is part of a series of three robot evolution simulations:
- WallerGlobal: Uses continuous movement and global perception for wall-following
- WallerLocal (this project): Focuses on ball-following with local perception
- WallerRestricted: Implements discrete movement in a smaller environment
Each variation explores different aspects of evolutionary robotics and navigation strategies.
This project is licensed under the MIT License. See the LICENSE file for details.
This software is provided "as is", without warranty of any kind. Use at your own risk.