Preprint / Version 1

Computer Vision Based AI for Control of High Velocity Autonomous Vehicle With Lane Detection Using Polynomial Regression


  • Jeffrey Aaron Jeyasingh Arcadia High School
  • Tony Smoragiewicz



Computer Vision , AI, High Velocity Autonomous Vehicle


Line following algorithms often fail to account for the po- sition of the vehicle within the lane, only depending on the curvature of the lane to determine steering output. For ex- ample, with previous algorithms, the vehicle could be to the right of the lane and the steering angle from the algorithm’s output would not correctly return the car to the center of the lane. Additionally, error can result from incorrect calibration: if the vehicle is truly pointing at a steering angle of 61 degrees when the correct angle should be 60 degrees, the vehicle will drift 1 degree throughout the turn. The error is accumulated and ultimately will cause the vehicle to leave the lane. By fitting a polynomial curve to the input image data, the exact location of the line can be determined relative to the vehicle. Furthermore, the center of the lane can be pinpointed, offset calculated, and trajectory modified to smoothly return to the center of the lane.


Bechtel, M. G.; McEllhiney, E.; Kim, M.; and Yun, H. 2018. DeepPicar: A Low-cost Deep Neural Network-based Au- tonomous Car. arXiv:1712.08644.

Bojarski, M.; Testa, D. D.; Dworakowski, D.; Firner, B.; Flepp, B.; Goyal, P.; Jackel, L. D.; Monfort, M.; Muller, U.; Zhang, J.; Zhang, X.; Zhao, J.; and Zieba, K. 2016. End to End Learning for Self-Driving Cars. arXiv:1604.07316.

Tian, D. 2019. DeepPiCar-part 5: Autonomous Lane Navi- gation via Deep Learning.