CAN Bus Anomaly Detection in Autonomous Vehicle Sensor Data

Overview

This project focuses on detecting anomalies in the CAN (Controller Area Network) bus data collected from sensors in autonomous vehicles. The dataset includes sensor readings such as accelerometer RMS values, current, pressure, temperature, thermocouple measurements, voltage, and volume flow rate, which can be used to identify irregular patterns or malfunctions.

The dataset contains time-series data from various vehicle sensors, which is critical for real-time anomaly detection in safety-critical systems like autonomous vehicles.

Dataset Description

Columns:

• tag: Unique identifier for each data point.
• datetime: Timestamp when the data was recorded.
• Accelerometer1RMS: RMS value of the first accelerometer.
• Accelerometer2RMS: RMS value of the second accelerometer.
• Current: Current measurement in the vehicle's system.
• Pressure: Pressure reading from the sensors.
• Temperature: Internal vehicle temperature (ambient).
• Thermocouple: Thermocouple temperature measurement.
• Voltage: Voltage reading from the system.
• Volume Flow RateRMS: RMS value of the volume flow rate.

Sample Data:

tag	datetime	Accelerometer1RMS	Accelerometer2RMS	Current	Pressure	Temperature	Thermocouple	Voltage	Volume Flow RateRMS
1	2/8/2020 13:30	0.20603	0.277924	1.81019	0.382638	90.174	26.776	228.208	121.664
2	2/8/2020 13:30	0.204366	0.275727	2.66317	-0.273216	90.2836	26.776	227.245	122
...	...	...	...	...	...	...	...	...	...

Objective

The primary goal is to develop a model that can identify anomalies in the sensor data, which could indicate hardware issues, environmental changes, or system failures. Anomalies are important to detect as they may lead to unsafe driving conditions in autonomous vehicles.

Key Steps

1. Data Preprocessing:

   • Handling missing values.
   • Scaling and normalizing sensor readings.
   • Feature engineering to extract additional insights from the raw data.

2. Exploratory Data Analysis:

   • Visualizing time-series data to understand normal vs. abnormal behavior.
   • Analyzing the distribution of various sensor readings.

3. Anomaly Detection Techniques:

   • Unsupervised Learning: Using clustering methods such as DBSCAN, Isolation Forest, or Autoencoders for detecting outliers in the data.
   • Supervised Learning: Training a model using labeled data to classify normal and anomalous patterns.

4. Model Evaluation:

   • Using precision, recall, F1-score, and AUC-ROC curves to evaluate the performance of the anomaly detection model.
   • Cross-validation and hyperparameter tuning for optimal results.

Requirements

• Python 3.8+
• Required libraries:
  • pandas
  • numpy
  • scikit-learn
  • matplotlib
  • seaborn
  • tensorflow/keras (for deep learning models)
  • pyod (Python Outlier Detection Library)

Installation

1. Clone this repository:

   git clone https://github.com/yourusername/can-bus-anomaly-detection.git
   cd can-bus-anomaly-detection

2. Install the required dependencies:

   pip install -r requirements.txt

Usage

1. Load the dataset and preprocess the data:

   from preprocessing import load_and_preprocess_data
   
   data = load_and_preprocess_data("data/can_bus_data.csv")

2. Train the anomaly detection model:

   from model import train_model
   
   model = train_model(data)

3. Visualize and analyze the results:

   from utils import plot_results
   
   plot_results(data, model)

Results

The trained model will highlight anomalies in the dataset, which can then be analyzed to understand potential malfunctions or irregular patterns in the vehicle's sensor system.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Special thanks to the contributors and the community working on autonomous vehicle systems, and the developers of open-source tools that made this project possible.