Autonomous Driving Using Imitation Learning

Published:

An imitation learning model that learns from an expert’s driving data and reproduces the expert’s commands in the same or similar situations.

  • Data were collected in the CARLA simulator to stand in for real-world driving.
  • The model is a ResNet-18 policy trained with behavior cloning.
  • It needs only a small amount of training data.

Tech: Python, PyTorch, CARLA, ResNet-18

Source code