carla_yaw

End-to-end Autonomous Driving System using yaw information.

We implement a vision-based autonomous driving mode using yaw and localization based End-to-end autonomous driving model with pytorch.

Implemented Version

• Python 3.6
• Pytorch 1.3
• Torchvision 0.2
• CARLA 0.8.2

Kinematic Vehicle Equation

We utilize the bicycle model form kinematic vehicle dynamics. The yaw angle is calculated by Kinematic Vehicle Equation.

How to Run

• Train: python main_auxi_v0.py --lr=1e-4 --train-dir=your dir --eval-dir=your dir
• Eval: python run_auxi_ver0.py --model-path=trained_model.pth

CARLA benchmark

The CARLA simulator has a large variety of driving environments, such as traff ic lights and dynamic obstacles, including dynamic vehicles and pedestrians.

You can download the CARLA simulator from here (benchmark).

licence : LGPL 2.1