Ethics Hub

This page contains information about ethics and data privacy in the RobotCycle project, with a particular focus on data privacy.

Project Goal

RobotCycle is a sponsored project at the intersection of Machine Learning, Robotics, Computer Vision, and urban development. This project aims to gain insights into how cycling infrastructure can affect cyclists' behaviour and safety during their journeys. Project outcomes will inform the decision-making of different stakeholders (including cyclists, other road users, city planners, and/or policymakers).

Overall, the project has the following objectives:

• Collect real-world traffic data capturing cyclists and other road users.
• Analyse and assess the road network in terms of safety for cyclists.
• Categorise road infrastructure according to safety and identify areas of high risk.  

Below, you can find a privacy notice and further information about ethics and data used in the project.

Privacy Notice

One of RobotCycle’s main deliverables is the collection of a real-world traffic dataset capturing cyclists and other road users through sensor recordings including images. This data will be used to analyse and assess Oxford’s road network in terms of safety for cyclists, categorise road infrastructure according to safety, and identify areas of high risk.

We will collect data from multiple sensors from a sensorised backpack while riding a bike in Oxford’s urban area. The sensors have been chosen to gather information about the road's different road users, traffic, and infrastructure.

The backpack will be clearly visible due to the sensors attached to its top. The sensors include four cameras, two laser sensors and one GPS sensor. In addition, we will record camera data from a camera attached to the rider’s head. Moreover, the rider will be clearly visible as they will wear high-visibility jacket and the backpack itself will be covered with a high-visibility sticker.

We selected three different routes for collecting data within Oxford. The three routes have been designed to avoid health clinics and locations associated with religious worship, locations of the Department of Engineering Science where we may be more likely to recognise individuals, and, where possible, college locations. The detailed routes can be consulted at https://www.google.com/maps/d/viewer?mid=1AhKpvLpzUm2D1mruWtTo0y0dijjdV-s&ll=51.76699156418116%2C-1.2602713534118615&z=15.

Personal information (e.g., faces) may be recorded incidentally in images and videos. In RobotCycle, we consider datasets collected to represent generic objects and situations. The datasets do not focus on specific individuals or other personal or sensitive information. In RobotCycle, no research is conducted on extracting biometrics such as face recognition. One research team member will blur collected data for faces, car number plates, and other identifiable information before it is disseminated to others for analysis.

The University of Oxford is the data controller for your personal data, which will determine how your personal data is used in the research. The University will process your personal data for the research outlined above. Research is a task that is performed in the public interest. Further information about your rights with respect to your personal data is available from the University’s Information Compliance web site at https://compliance.admin.ox.ac.uk/individual-rights.

How is the data processed?

The data in RobotCycle is used for non-commercial scientific research purposes. Sensor data – including videos and laser data -- are used to develop algorithms to automatically extract useful information from images, videos and point clouds. These algorithms are implemented in computer programs.

In RobotCycle, algorithms focus on the specific tasks of mapping and understanding the environment. For instance, a 3D reconstruction of a road in Oxford can be created, and the objects and road users can be detected there. Given the road layout, the main application is understanding or predicting risky situations.

The algorithms in RobotCycle are not designed to extract biometrics or other sensitive characteristics, even when these could be potentially inferred from images and videos. This means that while some algorithms can tell whether a person is in an image or their pose (i.e., the location of the limbs in 2D or 3D space), they cannot recognize the person's identity or characteristics.

Using this research data is necessary for public research in this space. The data is securely stored on Oxford premises (data centre), access to it is restricted, and it will be retained as long as it is deemed of value for the community.

Why is the data processed?

Collecting a custom dataset is necessary to allow research to be conducted in the area of the project. We need a custom dataset from a cyclist's point of view to traverse the environment as close as possible as the subjects of our study would do. This will give us the possibility to gather information in an innovative way, as most datasets are taken from vehicles. We selected the sensors that would give us the best information for this task, including cameras and laser sensors. Together with the uncommon point of view, these will ensure our ability to map the environment regarding safety for vulnerable road users.

Who is using the data?

The data is used by researchers affiliated with the RobotCycle project and, more broadly, with the Oxford Robotics Institute. In addition to the PI, these are usually doctoral students, postdoctoral researchers, and technical staff sponsored by the project.

In some cases, other researchers in our laboratory, universities, or research institutions may collaborate on parts of the research. Moreover, the dataset in its entirety of sensor recordings will be made available for research institutions after authorisation, which is granted if, and only if, the user downloading the data has an active institutional email address. In any case, only the masked data – i.e. where possible sources of identification have been blurred/deleted – will be made available for research.

This project will be part of the academic dissemination into conference papers, journal articles, and relative presentations at national and international conferences. Similarly, part of this project will be included in student theses. The dataset will be made public to academic institutions internationally, allowing research to be conducted. As such, details of this project will be presented in papers, articles, and theses at other institutions.

Results and discussions on this work will be presented nationally and internationally at invited talks at conferences or to industrial partners. Similarly, materials (videos, flyers, blog posts, etc.) will be produced to promote the project and the Oxford Robotics Institute (ORI) in general. Finally, we intend to continue this line of work through additional funding. Thus, we will use the results from this project to apply to grant calls and engage with possible stakeholders or private investors.

What data is shared?

Only a dedicated engineer and a system administrator will have access to the raw data. They will mask all the data to eliminate all the information that could lead to identification – e.g. human faces and car plates. This process will be carried out on an internal server on the internal network. The masking process will be based on state-of-the-art machine learning algorithms, which we will run locally on the server.

Data will be made available to the community through a download service, allowing only users with active academic credentials to download it. The service will keep information on these users and ask them to delete/redownload the data if any modifications/deletions happen. These data will include all sensor recordings – where potentially identifiable information will be blurred/masked.

How to request data erasure?

If you think your personal information is used in the project and wish it to be erased, please send us an email following these instructions:

• Send an email with this information to the project’s webmail, robotcycle@oxfordrobotics.institue
• In the email subject line, please enter Data Erasure Request.
• Clearly specify which image(s) or video(s) you would like us to erase
• State a legitimate reason for your request (e.g., the image or video contains your face)

After verifying your request, we will delete the relevant image(s) or video(s) from our system and stop using them in the research. We will also notify all the institutions that downloaded our data to request deletion on their side.

Data retention

Identifiable data will be stored in secure, encrypted, and local storage for as long as it continues to be valuable to the research community.