This repository is meant to control a car model, using neural networks in order to follow a trajectory within his lane and avoid moving obstacles.
Watch it in action in the following video
This network was inspired by the residual neural network architecture, specially the densenet approach to concatenate the outputs of the skip layer connections instead of adding them, using just 48,300 parameters.
- ROS (melodic, 1.14.3)
- Ubuntu (16.04.5 LTS)
- python-catkin-tools
On the root directory (where this README.md
file can be found) execute the following commands:
catkin build
source devel/setup.bash
Currently the simulation comes from another project, so please follow the instructions on the the next repository's wiki in order to set the car model simulator.
Then open a terminal on the root directory and execute the following commands:
roslaunch ai_control.launch