Skip to content

Latest commit

 

History

History
40 lines (27 loc) · 2.4 KB

README.md

File metadata and controls

40 lines (27 loc) · 2.4 KB

Visual Task Specification for Lane Control

A highway self-driving model that uses Visual Task Specifications primarily seen in robotics and applies them to the output of a deep learning lane detection model to find steering and throttle inputs to keep the vehicle in the lane. The model was tested in a simulated setting in Grand Theft Auto V's highways.

More information on visual tasks specifications can be found in this paper by Gridseth et. al., or this paper by Hespanha et. al.

Requirements

Install the required packages using the following command:

pip install -r requirements.txt

You will also need the following programs/packages:

x360ce

vJoy

• Grand Theft Auto V; optionally install Simple Trainer for GTA V to control weather, time, other vehicles and more to fine tune your testing setting.

Instructions

  1. Install required packages and programs
  2. Download the TuSimple model from here and place the .onnx file into the models folder
  3. Run x360ce, vJoy, and Grand Theft Auto V
  4. Place character in a "Brute Camper" vehicle on the highway
  5. Run lanecontrol.py
  6. Change driving settings with the control panel

Results

Acknowledgements

• the pyvjoy directory was obtained from the github repository here

• the ultrafastLaneDetector directory was obtained from the github repository here

• the models directory was obtained from the github repository here

• the grab_screen() function in the grabscreen.py file was obtained from the github repository here

• all other python files and functions are original and written by Jeramy Luo