Autopilot system for Tesla cars

Mentors: Yeeshukant Singh, Kshitij Bhat

Members :

• Aditya Suwalka

• Tarun Kumar Patel

• Soham Mondal

• Anurag Srivastava

Description:

• While on the highway, the monotonous speed control is a bit tiresome for drivers (especially for long distances and longer durations). Lane Assist and Adaptive Cruise Control (ACC) features can help the drivers in these situations. Depending on the actions of other objects/vehicles in the car's immediate vicinity, these systems can slow down and stop the vehicle when required.

Specifications:

• Computer Vision Pipeline for Lane detection should be efficient in detecting the free path on the road.
• Motion planning and control systems are needed to steer the vehicle to have the least cross-track error and automatic lane centering.
• Bonus points for an active obstacle avoidance system to change lanes and avoid other cars in the lane.

DelyBot

Ready to serve you anytime, anywhere.

About The Project

Built With

• ROS Noetic
• ROS Navigation Stack
• ROS robot_localization
• Gazebo
• Python

(back to top)

Getting Started

Prerequisites

At first it is necessary to install external packages:

• navigation

  sudo apt install ros-noetic-navigation

• slam-gmapping

  sudo apt install ros-noetic-slam-gmapping

• map-server

  sudo apt install ros-noetic-map-server

• robot_localization

  sudo apt-get install ros-noetic-robot-localization

HOW IT WORKS

Robot navigation is implemented in the delybot_navigation package.

In the amcl.launch file, Adaptive Monte Carlo Localization (AMCL) is implemented, which is a method for localizing the robot and tracking its pose in a map. The optimal path planning from the robot pose to the target pose with prior knowledge of the environment and static obstacles is done by the basic Global Planner. On the other hand, the Local Planner computes in real-time the new path to avoid dynamic obstacles. It is possible to choose between two implemented Local Planners: DWA and Trajectory Rollout, both of which work efficiently.

Installation

1. First make a catkin workspace

   mkdir -p catkin_ws/src
   

2. Clone the repo inside the catkin workspace

   cd ~/catkin_ws/src
   git clone https://github.com/git-suwalkaaditya/DelyBot.git

3. Build packages

   cd ~/catkin_ws
   catkin_make

4. Before running any roslaunch and rosrun command, run the following command

   source devel/setup.bash

(back to top)

Usage

The possible optional parameter values are listed in the table.

Optional Parameter	Values	Default
open_rviz	true, false	true
world	empty, delybot_test_map, delybot_test, district_map, district	district
dwa_local_planner	true, false	true

The world parameter can be used to load a specific world map (the .world file must be located in the delybot_description/world/ folder).
To test the different world files in Gazebo it is possible to run the following command:

# Usage example to open district.world in Gazebo
roslaunch gazebo_ros empty_world world_name:=/home/your_username/catkin_ws/src/DelyBot/delybot_description/world/district.world

delybot_description

roslaunch delybot_description display.launch

Open a pre-configured Rviz session to display the robot.

roslaunch delybot_description gazebo.launch open_rviz:=true

Spawn the robot into the gazebo simulation environment. If the open_rviz optional parameter is true a pre-configured Rviz session is also opened.

delybot_control

roslaunch delybot_control ddr_control.launch world:=district

Spawn the robot with a differential drive control and a teleoperation node in gazebo.

delybot_slam

roslaunch delybot_slam delybot_slam.launch world:=district_map

This command can be useful to create a 2d map of a specific world using a gmapping algorithm.

delybot_navigation

roslaunch delybot_navigation delybot_navigation.launch world:=district dwa_local_planner:=true

This command is used for the robot navigation, it's possible to give through Rviz a desired goal pose.
If the dwa_local_planner parameter is true the DWA local planner is used, else if it is false the Trajectory Rollout local planner is used.

waypoint_spawner node

rosrun delybot_navigation waypoint_spawner.py

This command run the waypoint_spawner node, used to send a specific goal pose selected from a predefined list to the robot.
The list is imported from the waypoint.json file inside the delybot_navigation/scripts/ folder.

# Output example
user@user:~$ rosrun delybot_navigation waypoint_spawner.py 

GOAL LIST:
0) Origin
1) Thrift Shop
2) Salon
3) Home
4) Post Office
5) Police
6) School
7) Fast Food

Insert goal index: 

Roadmap

• 3D robot modeling
• Add differential drive control
• Add laser and imu sensors using Kalman filter (pkg: robot_localization)
• Add world 3D model with static and moving obstacle
• Add delybot slam using gmapping to create 2D world map
• Add delybot navigation
• Add waypoint spawner node
• Update README

(back to top)

Gazebo Simulation

https://github.com/git-suwalkaaditya/IITISoC-22-Auto4-Autopilot-System-Cars/blob/main/simulation%20video.webm

(back to top)

Reference

For further information check the report - DelyBot Report.pdf

https://roscon.ros.org/2017/presentations/ROSCon%202017%20Vehicle%20and%20City%20Simulation.pdf

https://github.com/theAIGuysCode/YOLOv4-Cloud-Tutorial

https://github.com/osrf/car_demo

https://github.com/gmeidk/DelyBot

http://kth.diva-portal.org/smash/get/diva2:1057261/FULLTEXT01.pdf

https://medium.com/intro-to-artificial-intelligence/model-predictive-control-udacitys-self-driving-car-nanodegree-ad7cf64fd0e4

[1] navigation package: http://wiki.ros.org/navigation

[2] slam-gmapping package: http://wiki.ros.org/gmapping

[3] map-server package: http://wiki.ros.org/map_server

[4] robot-localization package: http://wiki.ros.org/robot_localization

[5] actor: http://gazebosim.org/tutorials?tut=actor&cat=build_robo

(back to top)

Contacts

Aditya Suwalka:

Tarun Kumar Patel

Soham Mondal:

Anurag Srivastava:

TAKE A LOOK AT BRANCH 3 ALSO