ShaneXiangH/Application-Level-Anomaly-Detection-in-Autonomous-Vehicle

A collision anomaly detection using Carla

Application-Level-Anomaly-Detection-in-Autonomous-Vehicle

This work is done in 2020 UCInspire program.

introduction

Autonomous vehicle is one of the development directions in future society. One topic autonomous vehicle research should take great care of is the security. Accident should be completely avoided from happen on autonomous vehicles. To achieve this goal, an autonomous vehicle is supposed to be equipped with reliable anomaly detection mechanism. In general, to do the anomaly detection, an autonomous vehicle needs some related sensors to listen and collect real time sensor data. In addition, it needs one or some efficient algorithms to detect abnormal situation. In this project, I use Carla simulator to do the real time vehicle driving simulation, listening and collecting data from different kinds of sensors, and using outlier detection algorithms to detect abnormal situation.

The whole experiment process:

The final diagram is like this:

The projects used are:

Carla Simulator

Python Outlier Detection (PyOD)

Darknet Yolo v4

The way to run the project

1. Clone the Carla Simulator, Python Outlier Detection (PyOD), and Darknet Yolo v4 separately in your PC, add the files in this repository with the right path.
2. Run the Carla Simulator, Yolo v4, and PyOD based on the whole experiment process