Pedestrain Avoidance

There are three parts including simulation in Gazebo, people tracking and control of the vehicle. Youtube link: https://www.youtube.com/watch?v=-01lbQc_njY&t=5s

People Tracking: Using CNN to detect person. Kalman Filter and Matching algorithm to track people, based on SORT algorithm. Using RGBD camera and 1-layer Lidar to perceive environment.

Collision Avoidance: Treate it as a optimization problem. MPC + Linearize Bicycle Model + Convexify collision avoidance constraints(sequential convex method).

Mobile Robot

Simulation in Gazebo

Code Structure

actor_plugin and vehicle_plugin: Interface between ROS and GAZEBO which send command to objects in GAZEBO and send data from GAZEBO to ROS.

my_car: Model files of vehicle, sensors(RGBD camera and LIDAR) and the world file.

people_sim: Model files of pedestrains.

rgbd_detector: Object Detection and tracking people.

collision_avoidance: motion planning and control.

Prerequisites

ROS
GAZEBO
Eigen
OpenCV
CPLEX

Installing

1. Clone this reporsitory in workspace.
2. Modify YOLO Library path in CmakeList of rgbd_detector.
3. Modify CPLEX path in CmakeList of avoidance.
4. catkin_make in this workspace
5. Move libactorplugin_ros.so and libmy_controller_lib.so generated in actor_plugin and vehicle_plugin into the directory ~/ws/devel/lib/ 
6. roslaunch al.launch

Maintainer

Jieming Chen(15001781672@163.com)

Hao Li

References

[1] Bewley, Alex, et al. "Simple online and realtime tracking." IEEE, 2016.

[2] Redmon, Joseph, and Ali Farhadi. "Yolov3: An incremental improvement."

[3] Ribeiro, Maria Isabel, and Pedro Lima. "Kinematics models of mobile robots."

[4] Alrifaee, Bassam, Janis Maczijewski, and Dirk Abel. "Sequential convex programming MPC for dynamic vehicle collision avoidance."