Internship | Traffic Light Recognition Perception System Development

One of the challenges in Automated Driving is “Environment Perception”. The vehicle decision is based on the current assessment of the vehicles’ surroundings. In order to do so, different environment perception algorithms are required. One of the systems that could help autonomous vehicles increase traffic flow and throughput is Green Light Optimized Speed Assist (GLOSA). For GLOSA, the vehicle receives information about the current and next phase of the traffic light status, including switching time. With onboard perception, the vehicle needs to detect the traffic light status in order to confirm the correctness of the received traffic light information.

In this internship you will:

• Train an algorithm for automated Traffic Light Recognition
• Integrate the traffic light recognition on one of TNO’s research vehicles and validate the performance
• Develop tools for visualization of the traffic light status in a graphical user interface

Proposed approach:

• Data Preparation: Inspect the in-house, annotated dataset with traffic light 2D bounding boxes, ensure proper preprocessing of the data for training.
• Model Selection and Development: Choose or customize a neural network architecture using frameworks like TensorFlow or PyTorch, tailoring it for the target perception task of real-time object detection.
• Training the Model: Train the model on the dataset and include hyperparameter tuning. The output format is still to be determined, but it will most probably be 2D bounding box estimation, with association to the lane network (i.e., which traffic light corresponds to which lane?).
• Documentation: Document the system architecture, datasets, and testing results.
• When time allows:
  • Integration with Vehicle Systems: Deploy the trained model on an onboard computing platform (e.g., Nvidia Jetson), ensuring compatibility with real-time camera input and processing pipelines.
  • Real-World Testing and Iteration: Test the system in diverse operational scenarios, gather data on edge cases, and iteratively improve the model by retraining with additional data.

What can you expect of your work situation?

You will work at the Integrated Vehicle Safety department of TNO on the Automotive Campus in Helmond. In this department, people are working on developing software for automated driving vehicles. The developed software is tested in pilots and on the public road. The people are young, enthusiastic, and driven. You will work in an open area, within your own team. One of our employees will be your mentor. He will help you to get acquainted with the department and give you guidelines for your research in order to help you to get the best out of it.