OPA3L is the short version of the project name “Optimal Assisted, highly Automated, Autonomous and Cooperative Vehicle Navigation and Localization”. Engineering mathematicians (Optimization and Optimal Control working group) and also computer scientists (Cognitive Neuroinformatics working group and Virtual reality working group) from the University of Bremen wish to apply integrated approaches for autonomous and cooperative driving in the frame of the project. The project will run for four years. DLR Space Administration and the German Federal Ministry for Economic Affairs and Energy (BMWi) are providing 4.5 million euros of the total project volume of 5.3 million euros. The involved industry partners are funding a further 800,000 euros. It is intended that 2.5 million euros of the project volume go towards research at the University of Bremen.
Autonomous Driving: Many Visions, Many Expectations
OPA3L is the continuation of the AO-Car project (“Autonome, optimale Fahrzeugnavigation und -steuerung im Fahrzeug-Fahrgast-Nahbereich für den städtischen Bereich”) (Autonomous, Optimum Vehicle Navigation and Controlling in the Vehicle-Passenger-Area for urban areas). Said project was conducted at the University of Bremen in the last two years. “The development of self-driving cars is extremely current and is widely discussed. There are many visions, but also many expectations”, stated Professor Christof Büskens from the Center for Industrial Mathematics (ZeTeM) at the University of Bremen. “We want to attain sustainable, needs-based and efficient mobility, less status, more safety and comfort. However, the integration of autonomous vehicles in the traffic system, which has been around for decades and was designed with people in mind, poses a huge challenge.”
“Element of uncertainty: People”
The Bremen researchers have committed themselves to these challenges. “The vehicle must be able to orient itself in limited and strongly rule-based environments, must make long-sighted and optimum decisions and must react to the complex, unforeseeable and biggest element of uncertainty: people”, explained the project manager Dr.-Ing. Mitja Echim. In the future, the communication between the vehicles and their environment will play an important role in pointing out dangerous situations or carrying out optimum and cooperative driving manoeuvres.
From the University Parking Lot to the City Roads
In the previous project AO-Car, the research vehicle was able to navigate autonomously around the university parking lot. Obstacles were taken into account and free parking spaces were detected so that a subsequent, fully autonomous parking manoeuvre could be carried out. The researchers will now go one step further: They are venturing into urban areas. A particular focus is placed on situations that are characterized by an immediate use in day-to-day life. Autonomous shuttle services to the next tram stop in city outskirts with an insufficient public transport network or the autonomous provision of vehicles from car sharing services right on your doorstep are examples of this.
“The challenge is to determine current vehicle situations based on multi-sensory information, prior knowledge and real-time information”, defined Dr. Joachim Clemens from the Cognitive Neuroinformatics working group. “At the same time, the environment must be mapped. Depending on the situation, the right manoeuvre with the optimum control and regulations needs to be calculated.” Risk assessments, which consider the available “uncertain knowledge”, are necessary. This in turn can lead to changes in behavior and the generation of new manoeuvres.
Cooperative and linked manoeuvres, such as environmental and economical right-of-way rules at crossings, the joint starting-up at traffic lights, cooperative usage of several lanes or the communication with oncoming vehicles on particularly small streets will be of particular focus. Two new industry partners that have just been integrated into the OPA3L project will provide three further autonomous vehicles and two additional vehicles, which will be controlled by humans, will be fitted with sensors and communication equipment.
From Space to Earth
The involved working groups already yielded considerable results in the previous AO-Car project. The development of the main driving functions was carried out in the short period of less than a year. This was possible due to the circumstance that team transferred the methods and algorithms developed by them for space applications, such as the autonomous deep-space satellite navigation, to the terrestrial field of “autonomous driving”.
These challenging concepts from the field of artificial intelligence were applied to a hybrid production vehicle, which has only been changed insignificantly. Alongside the already significant standard equipment, the research vehicle, which will also be used for the new OPA3L project, is fitted with two additional laser scanners for the measurement of area distance at the front and rear of the vehicle and a GNSS+ system for highly specific, global localization. What was enough for the parking lot must now be adapted to the road.
In order to fulfil the requirements for reliable perception whilst driving in the city, the equipment will be expanded by means of additional laser sensors and cameras. However, before the vehicle goes on the road, the algorithms will undergo a multistep validation process in the frame of the project. Firstly, an elaborate real time simulation system will be used, which shows the exact behavior of the real vehicle on the computer in the form of a so-called ‘digital twin’. This will allow for both comprehensive day-to-day driving manoeuvres and extremely rare, extreme situations to be simulated and analyzed on a computer.
Failures Simulated with Model Cars
Secondly, several model cars at a scale of 1:8 will be used. Their sensors and actuator equipment are comparable to those of a ‘real’ vehicle. Unforeseeable situations caused by external disruptions will be simulated with them. The subsequent behavior – how does the vehicle react when a tree suddenly falls onto the road? – will be analyzed with the models. This level will offer University of Bremen students the chance to partake in current scientific problems in research at an early stage in their degree.
“Only afterwards will the algorithms that have been tested comprehensively in this way be investigated with a real car – initially without any other road users and then with increasing numbers of road users”, explained Christof Büskens. “In order to guarantee maximum safety, we continue to work with drivers and passengers. One person can take over control of the car at any time and the other person can continually check the system software. Both driver and passenger have been especially trained for this.”
www.math.uni-bremen.de/zetem/ao-car (in German only)
Prof. Dr. Christof Büskens
Optimization and Optimal Control
Center for Industrial Mathematics (ZeTeM)
University of Bremen
Faculty of Mathematics / Computer Science
Phone: +49 421 218-63861
Email: bueskensprotect me ?!math.uni-bremenprotect me ?!.de