for Connected Autonomous Driving using Deep Neural Networks and Multi-Sensor Fusion

Introduction:

Connected autonomous driving has become one of the most rapidly advancing areas of research in the field of computer vision and machine learning. Object detection and tracking are two key components of an autonomous driving system.
In recent years, deep learning techniques such as Convolutional Neural Networks (CNNs) have achieved state-of-the-art performance in object detection and tracking. Multi-sensor fusion, where data from multiple sensors are combined, can improve the robustness and reliability of object detection and tracking. This thesis proposes a camera-based object detection and tracking system that employs deep neural networks and multi-sensor fusion for connected autonomous driving.

Research Objectives:

The primary objective of this research is to design and implement a camera-based object detection and tracking system for connected autonomous driving that uses deep neural networks and multi- sensor fusion.

Specifically, the objectives of the research are:

• To develop a deep neural network-based object detection and tracking system using camera- based sensors.
• To investigate the use of multi-sensor fusion techniques to improve the robustness and reliability of the object detection and tracking system.
• To evaluate the performance of the proposed system using real-world driving data and compare it with existing camera-based systems.
• To assess the feasibility of deploying the proposed system on a connected autonomous vehicle platform.
  

Methodology:

The proposed research will employ a combination of deep learning, computer vision, and multi- sensor fusion techniques. The methodology will consist of the following steps:

Data collection: A dataset of real-world driving scenarios will be collected using a camera-based sensor system. The dataset will include diverse driving scenarios such as urban and highway.

Object detection and tracking: A deep neural network-based object detection and tracking system will be developed using the collected data. The system will use state-of-the-art object detection and tracking algorithms.

Multi-sensor fusion: The object detection and tracking system will be extended to include data from multi cameras. The data from these sensors will be fused using techniques such as Kalman filtering.

Performance evaluation: The proposed system will be evaluated using the collected dataset and compared with existing solutions. The performance metrics will include accuracy, precision and real time performance.

Feasibility assessment: The feasibility of deploying the proposed system on a connected autonomous vehicle platform will be assessed. The assessment will consider factors such as computational complexity, real-time performance, and system integration.

Expected Results:

The expected results of this research are:

• A deep neural network-based object detection and tracking system using camera-basedsensors for connected autonomous driving.
• An improved object detection and tracking system using multi-sensor fusion techniques.
• An evaluation of the proposed system using real-world driving data and comparison withexisting camera-based systems.
• An assessment of the feasibility of deploying the proposed system on a connected autonomous vehicle platform.

Conclusion:

This research proposes a camera-based object detection and tracking system that employs deep neural networks and multi-sensor fusion for connected autonomous driving. The proposed system has the potential to improve the reliability and robustness of object detection and tracking in adverse weather and lighting conditions. The research will contribute to the development of autonomous driving technology and has practical applications in areas such as transportation, logistics, and public safety.

Duration: 6 month 
Workplace: Science Park Graz 
Start Date: immediate

If you are interested, please send your application including your CV to:

Thomas Strasser-Krauss
thomas@tomrobotics.at