We made this car for our ELEC 424 final project at Rice University. The car autonomously steers to stay centered along the track (two lines of blue tape), and stops at stop signs (red squares on the track) using computer vision. The image processing and motor control is done using a BeagleBone AI-64. This project is based on Autonomous Lane-Keeping Car Using RaspberryPi and OpenCV.
To determine the resolution, proportional gain, and integral gain to use for the car we ran a series of trials and errors. By varying the gain and resolution by a slight amount after each attempt, we could tune the acceleration of the car until it was sufficiently accurate. For resolution, we found 160x 120 since it was sufficient and processed quickly.
We handled the stop box by detecting a large amount of red in the frame. We convert to HSV, and if the frame contains a large amount of red, then we stop.
1 / 3 • Plot showing error, proportional response, and derivative response vs frame number for the same entire run of the track.
1 / 2 • Picture of one of the vehicles with our hardware on it during the first few runs.
Video of a vehicle completing the course with your hardware and software running on the vehicle.
Comments