This is a multi-year project undertaken by the Cornell Data Science team intended to demonstrate autopilot for a full-size car.
CDS Self-Driving Car aims to demonstrate tight integration of a camera based vision algorithm to navigate a car safely. We utilize a SLAM system to localize ourselves, and an additional lane recognition pipeline to see the road. This is all built upon a Python control loop and will leverage robotics to directly actuate the steering wheel to augment a normal car.
This ambitious project was undertaken by Evan Williams, Iris Li, Edward Gu, Tobi Alade, Elias Little, Eric Zhang, Adarsh Sriram, and Shihao Cao
FINAL PRESENTATION: https://docs.google.com/presentation/d/1MIJ480jFvm6mGaNohGM7KqMZyA3pAJ67wQSmTtvUBTY/edit?usp=sharing
Here's some extra details about each of the individual parts that we developed and how they work.
https://github.com/maunesh/advanced-lane-detection-for-self-driving-cars
The main control loop framework was based on a Read-Estimate-Control-Command-Actuate architecture inspired by work from the Pathfinder for Autonomous Navigation team.
Blah blah
Please scroll to the bottom for continuation details.
See INSTALL.md
.
Always remember to have the virtual environment venv
activated with:
. venv/bin/activate
Our code is structured as a module: src
we invoke it with for example:
python -m src hootl.yaml NoSIM
This means to use the hootl.yaml
configuration file in src/configs
, along with the
NoSIM
testcase under runner/
. You can go to those directories respectively to find more config files
and more test cases.
TODO insert everything required to do an in car demo.