Prediction of human intention in focus in two new pre-studies

Dec, 13 2022

Being able to predict human behavior or ability in traffic may lead to positive traffic safety advantages. By knowing a road user’s next step, a warning to an approaching danger can be alerted or, in a vehicle, a safety system can be activated. Within the framework of the AI Enhanced Mobility program, in which SAFER is leading the traffic safety focus group, two new pre-studies will start up shortly to build knowledge in this field.

Malin Levin, project leader for the traffic safety focus group tells more:

“In the program's first call, both our submitted research ideas from the traffic safety group were granted funding. Understanding the intention of nearby road users, for example at a bus stop or a space with many different road users and transport modes is an important area for us, and we see that AI may have an important role to play. Our thesis is that we can help save lives in traffic by using image processing technologies, supported by AI algorithms, to enhance direct vision of for example bus drivers, remove blind spots, and detect and warn road users in danger”.

Intention Recognition of Vulnerable Road Users
In this project RISE, together with Volvo Buses, the Electricity project, Viscando and Univrses are aiming to understanding and communicating the intentions of road users such as pedestrians, cyclists, and scooter drivers around bus stops. Automatic recognition of intent is a key factor in tracking users and predicting trajectories in the near-term future. Such predictions can be used both in ADAS systems for human drivers, where these can alarm or initiate collision avoidance measures, and as an integral part in the control architecture of automated vehicles. For example, a bus driver could get help to recognize if a road user would like to cross in front of the bus, or if a pedestrian would like to get on the bus. 

The goal of pre-study is to conduct real-world data collection using ElectriCity Buses in Gothenburg and testing applicability of existing models for this data. The researchers are looking into developing edge solutions compliant with GDPR and implementing intention recognition algorithms in a real world setting. The pre-study is a stepping stone toward a larger project and helps to assert feasibility before starting other activities e.g. within the FFI-programme.

The project will start in December and will be concluded in April 2023. The ElectriCity project and Volvo Buses will provid the testbed, Viscando and Univrses provide the necessary sensors, and RISE will provide analytics solutions.

Improve traffic safety through advanced and automatic driving evaluations using AI and eye tracking
Today, about 60% of practical driving tests in Sweden result in failure. Also, about 25% of the population suffer from health conditions that can affect driving skills and require medical driving evaluations, but there is no way to do this in a systematic manner. The queues are long, and it usually takes about 8-9 months to get the evaluations done. During this time the driver may be at a higher risk for causing an incident. If it would be possible to get automatic driving evaluations using AI and eye tracking, this can fundamentally change the way the driving evaluations are conducted, thus facilitating new ways of conducting evaluations without the need of an expert in the vehicle. 

In the pre-study, researchers will investigate how eye-tracking in combination with AI could be used to, for example, see if a student is ready for a practical driving test or if a certain health condition would make a person less suitable to convey a vehicle safely. The researchers will look at whether tests could be automated, for example in a simulator. Using eye tracking can hence support our understanding about how we can predict intentions and navigate decisions. To conclude, improved driving evaluation methods can potentially change the way the driving evaluations are conducted, thus improving the traffic safety.

This pre-study will be carried by Örebro University and other stakeholders (RISE, VTI, Region Örebro Län (hospital), QTPIE and two driving schools in Örebro) will be part of a reference group to support the project. The activities within the project will be concluded during Q1, 2023.

Bringing actors together to utilize the potential
AI Enhanced Mobility is an important catalyst for accelerating the development of AI solutions in the field of mobility. The project aims to create new collaborations and projects, increases AI knowledge in the field of mobility and create synergies with existing initiatives and resources. The project is based on active participation from all partners, a needs-driven approach and cross-linking of AI and mobility competences. In total, more than 30 partners are engaged in the project.

SAFER leads the traffic safety focus group, and this strategic project focuses on bridging the knowledge gap that exists between needs owners and AI experts.  The overall goal of AI-driven mobility is to build experience, knowledge and create new collaborations in applied AI within the mobility sector to create conditions and solutions for sustainable mobility systems of the future.