Internship Data completeness assessment for vulnerable road user motion prediction

Internship Data completeness assessment for vulnerable road user motion prediction

Automated vehicles (AVs) have become widely popular in recent years since they have the potential to bring several benefits such as increased road safety, traffic throughput and fuel efficiency [1]–[3]. To achieve higher levels of automation (i.e., levels 4 and 5 of the SAE scale [4]), an AV must be capable of anticipating the actions of other road users in the driving environment. This is especially important for vulnerable road users (VRUs), such as pedestrians and cyclists, which accounted for almost 30% of the road casualties in the EU in 2017 [5]. Several types of models exist for this task, but data/pattern-based models have been a research focus in recent years [6] due to the success of deep learning in a various fields. However, a well-known challenge of these methods is the amount and quality of data required to train them successfully. Nevertheless, many state of the art methods for pedestrian motion prediction are trained and evaluated using traditional datasets (i.e., ETH [7] from 2009 and UCY [8] from 2007) which only contain a few hundred trajectories, despite the existence various recent and more complete datasets (e.g., [9]–[14]). Many other datasets exist, and so far, there does not exist a clear overview of the type of behavior that can be found in each of them. This is not an easy task, since the datasets are typically recorded at different locations, frequencies, times of the day… etc. Understanding the commonalities and differences in the interactions captured by each dataset could aid in overcoming the limitations of individual datasets and enable the development of better prediction models that have been exposed to all types of interactions during training.

What will be your role?

Automated vehicles (AVs) need to safely deal with vulnerable road users (VRUs), such as pedestrians and cyclists. Datasets are used to extract possible trajectories of VRUs, which can then be used to train prediction algorithms or to test the performance of AVs. Are you interested in analyzing the data and the pedestrian behavior that is found in the data? Please read further!

During this assignment you will be investigating the following research questions

• What kind of behavior and interactions can be found in different datasets? Is this behavior complementary, or similar across all datasets?
• What type of behavior is missing?
• Are there significant behavioral differences between different people (e.g. age group, gender, region)
• Can the different types of interactions be detected automatically?
• Can the prediction accuracy of data/pattern-based models benefit from complementing missing interactions of some datasets with behavior found in others?

Your assignment
To answer the research questions, you will:

• Focus on pedestrians first, then cyclists
• Perform a short literature review & make an overview of existing datasets that can be used for VRU motion prediction
• Perform a thorough analysis of the types of interactions and maneuvers recorded in the different datasets
• Implement at least 2 state of the art prediction models (may be taken from open-source repositories) and analyze their performance on the chosen datasets
• Enrich the chosen datasets with missing behavior that can be found in other datasets
• Analyze the performance of the prediction models when trained on the enriched datasets
• Give a final presentation & write a report, preferably in short article format

What we expect from you

• Good communication in English [Required]
• Experience with Python, Matlab and/or C++. Python is preferred [Required]
• 8.0 average or higher [Required]
• Experience with other common Python libraries such as NumPy, matplotlib, scikit-learn and SciPy (or equivalent) [Very recommended]

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 and testing automated driving vehicles. 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.

What you'll get in return

You want to work on the precursor of your career; a work placement gives you an opportunity to take a good look at your prospective future employer. TNO goes a step further. It’s not just looking that interests us; you and your knowledge are essential to our innovation. That’s why we attach a great deal of value to your personal and professional development. You will, of course, be properly supervised during your work placement and be given the scope for you to get the best out of yourself. Naturally, we provide suitable work placement compensation.

TNO as an employer

At TNO, we innovate for a healthier, safer and more sustainable life. And for a strong economy. Since 1932, we have been making knowledge and technology available for the common good. We find each other in wonder and ingenuity. We are driven to push boundaries. There is all the space and support for your talent and ambition. You work with people who will challenge you: who inspire you and want to learn from you. Our state-of-the-art facilities are there to realize your vision. What you do at TNO matters: impact makes the difference. Because with every innovation you contribute to tomorrow’s world. 

The selection process

After the first CV selection, the application process will be conducted by the concerning department. TNO will provide a suitable internship agreement. If you have any questions about this vacancy, you can contact the contact person mentioned below.

Due to Covid-19 and the consequent uncertainties and restrictions, students who are not residing in the Netherlands may currently not be able to start an internship or graduation project at TNO.