Within the last years, the automotive industry has done intense research regarding early detection of potential accidents in order to increase the safety on roads. Investigations include several different approaches based on laser scanners as well as radar and video technologies, which require a line-of-sight between the moving vehicles. Other approaches get along without a line-of-sight, as it is the case for collision detection based on cooperative systems. Thereby, the position, velocity etc. are derived by GNSS positioning and exchanged between the vehicles to detect possible collisions in time.
Within the proposed project UPIC, the idea of cooperative systems for collision prediction and avoidance is pursued. The planned system will use Car2Car communication (802.11p standard) to share the state vectors of each car. To keep the system as simple and cheap as possible, the positioning is aimed to be based on low-cost single-frequency GNSS equipment and should be widely independent of additional infrastructure. Therefore, common approaches as RTK or DGNSS, both dependent on continuous data from reference station networks, will not be applied.
For the positioning, the GNSS method Precise Point Positioning (PPP) shall be utilized to reach the required accuracy of a few decimetres by using code and phase measurements. The precise orbit and clock data needed for PPP can be downloaded from the internet periodically. This download option is needed because no continuous data link can be guaranteed. Since the GNSS accuracy is significantly decreased in urban regions, GNSS shall be combined with wheel sensor data to overcome GNSS-critical regions. Wheel sensor data is already accessible in modern cars and will be considered within a tightly coupled integration approach. In addition, sophisticated kinematic (e.g. bicycle model) or dynamic models for the Kalman filter shall further improve the results.
The derived state vectors including position, velocity, heading etc. are used in a collision algorithm to predict the vehicles' trajectories and further to detect potential collisions. For the reliable prediction of the trajectories, precise lane-level road maps as well as the knowledge of the lane of each car are needed. Therefore, a lane-level map matching will be executed prior to the prediction step. The extended state vector including lane information is exchanged between the cars to enable the detection of potential hazard situations. In case of a detection, the collision probability, time to collision and severity (collision speed, angle and point of impact) will be calculated.
Being aware of the fact, that systems based on cameras or radar sensors already deliver very reliable results for collision prediction, within this project, the potential of a system comprised of GNSS, wheel sensors and precise map information shall be highlighted. The proposed system is expected to show better performance compared to camera and radar systems in case of no line-of-sight (e.g. in curves or at crossroads) or in case of larger distances and shall help to detect dangerous situations significantly earlier. It is not intended to run such a system isolated, but showing the huge potential, GNSS would contribute in an increased road safety by being part of a combined camera-radar-GNSS system. Thinking of the market potential for GNSS within the automotive field, this would give GNSS a huge push.
In summary, the proposed project aims at the development of an affordable collision detection system based on GNSS positioning, wheel sensors, map information and Car2Car communication. The system will on the one hand show benefits in situations where camera or radar based systems are limited because of no line-of-sight. On the other hand, a GNSS based lane-level accurate positioning system will also be needed for autonomous driving, which is an up-to-date research topic. The outcome of the project should be an expandable demonstrator for reliable real-time collision detection.
TU Graz - Graz University of Technology
- Dr. Steffan - Datentechnik Gesellschaft m.b.H.
- MAGNA STEYR Fahrzeugtechnik AG & Co KG
TU Graz - Graz University of Technology