Contains V-REP simulation scene file with line-following Ackerman car. The algorithm is implemented on Python using V-REP remote API.

Report can be found in ControlTheoryProject/documentation/Report.pdf file.

The script can be found in ControlTheoryProject/V-REP simulation/AckermannCar folder.

Line-following Ackermann car

To run the project, clone ControlTheoryProject/V-REP simulation/AckermannCar folder to your computer.

Check V-REP remote api settings (i.e. the port number in configuration file is the same, as specified in simulationEnvironment/connections.py). Additional information can be found here: http://www.coppeliarobotics.com/helpFiles/en/remoteApiServerSide.htm

Also check settings from following page: http://www.coppeliarobotics.com/helpFiles/en/remoteApiClientSide.htm

Launch V-REP application. Then run main.py file.

Handling car movement

An interface for moving the car and reading sensors data is provided in simulationEnvironment/AckermannCar.py.

You can use following methods of AckermannCar instance:

• read_vision_sensors_intensity() which returns intensity values read from left, middle and right vision sensor respectively
• read_proximity_sensors() which returns destination read from right, middle and left proximity sensor respectively
• set_rotation_angle(base_angle) for setting steering angle value (in radians)
• set_speed(speed) sets speed to front motors

For running a car you can use one of controllers, provided in car_controllers package or create the own one that would have implementation of following methods:

• set_driven_car(car)
• drive_car()

How to run different algorithms

In main.py file run_program() function uncomment one of the following lines:

• run_single_test() for running single test
• run_test_engine() for running bunch of tests with predefined range of tested parameters (PID controller's coefficients and car speed)
• run_neat_tests() for running tests with NEAT algorithm
• run_best_genome() for running simulation with neural network of the fittest genome from certain generation

Algorithms running options examples

run_single_test()

This function runs one test with PidController.

run_test_engine()

It runs several tests with PidController using all possible combinations of parameters from specified ranges.

run_neat_tests()

It runs NEAT algorithm, which selects suitable neural network for driving Ackermann car

run_best_genome()

It restores specified generation from checkpoint (created by NEAT algorithm), gets a neural network of the best genome and runs simulation with NeatController using this network

Controllers description

UniversalController

It is a simple controller for line-following car. Implementation of function drive_car() from this class is used by PidController and NeatController For using UniversalController you can specify:

• type of sensors used by controller for checking relative car position:
  • sensor_type_vision
  • sensor_type_proximity
• limit of simulation time (in sec) or not restrict it by setting time_limit = -1

PidController

For this type of controller you should provide:

• PID coefficients (kp, ki, kd)
• speed of the car (base_speed)
• sensor type(as in UniversalController)

NeatController

For this type of controller you should provide:

• neural network (net)
• sensor type(as in UniversalController)

Also you can set your own range of parameters resulted by net.