I equipped this model car with an autonomous steering system for the IronCar race, in which each car had to complete 3 timed laps.
- RaspberryPi 3b+
- RC car (T2M Pirate Shooter)
- PiCamera (wide angle)
- RaspberryPi PWM shield
- some voltage regulators
- wifi dongle
- relay module as an emergency stop
The car's onboard computer processes the camera feed and continuously adjusts the wheel angle to keep the car on track. It uses a combination of convolutional and sequential neural networks. I experimented with some data augmentation techniques such as adding artificial shadows to the images, and mirroring them.
The other on-board computer (it was an Arduino at first but it has been replaced by a PWM shield) is in charge of retrieving the predictions transmitting the corrected angle to the motors.
