Igor Doric

Statistical Behavior Modeling for Driver-Adaptive Precrash Systems

Precrash systems have the potential for preventing or mitigating the results of an accident. However, optimal precrash activation can be only achieved by a driver-individual parameterization of the activation function. In this paper, an adaptation model is proposed, which calculates a driver-adapted activation threshold for the considered precrash algorithm. The model analyzes past situations to calculate a driver-individual activation threshold that achieves a desired activation frequency. The advantage of the proposed model is that the distribution is estimated using a distribution model. This has the result that an activation threshold can be already determined using a small data set. In addition, the confidence interval that has to be considered is decreased. The proposed model was applied in a study with test subjects. Results of this paper confirm the usability of the model. In comparison with an empirical approach, the proposed model achieves a significantly lower threshold and, thus, a higher safety effect of the system.

Test methodology for rain influence on automotive surround sensors

Automotive safety systems aim to provide maximum protection to vehicle occupants and vulnerable road users. These safety features rely on data from surround sensors such as radar, lidar and camera, which provide detailed information about the environment of the vehicle. A minor error in the measurements of these sensors can lead to major injuries or death. Hence, the reliability and accuracy of these sensor systems is mandatory. The performance of surround sensors depends on their local environment, because of the attenuation of the ambient atmosphere. Environmental influences such as rain additionally affects the accuracy of sensor systems. Therefore, these sensors must be tested under various weather conditions. This paper presents a new test methodology for rain influence on automotive surround sensors. Therefore, a rain simulator was designed and validated. The proposed test methodology was applied to radar, lidar and camera sensors in an experimental setup.

Modeling and simulation of rain for the test of automotive sensor systems

This paper presents a new approach for the test of automotive sensor systems in rain. The approach is based on an indoor test method, which helps to save test kilometers and test effort. For the activation of safety systems detailed information about the vehicles environment is necessary. Laser scanners provide precise information about the environment and a high angular resolution in contrast to radar sensors [1]. The performance of laser scanners depends on their local environment, because of the attenuation of the ambient atmosphere, precipitation and on the reflectivity of objects. False measurements in the field of vehicle safety can result in severe injury or death, so high reliability is essential. For this purpose a theoretical model is developed in order to determine the sensor behavior. Subsequently, a rain simulator is constructed to validate the theoretical model. Furthermore the developed rain simulator is validated by comparison with real rain. Based on determined rain disturbances benchmark tests of different sensor systems and algorithm approaches can be performed.

A Novel Approach for Researching Crossing Behavior and Risk Acceptance: The Pedestrian Simulator

To integrate unpredictable human behavior in the assessment of active and passive pedestrian safety systems, we introduce a virtual reality (VR)-based pedestrian simulation system. The device uses the Xsens Motion Capture platform and can be used without additional infrastructure. To show the systems applicability for pedestrian behavior studies, we conducted a pilot study evaluating the degree of realism such a system can achieve in a typical unregulated pedestrian crossing scenario. Six participants had to estimate vehicle speeds and distances in four scenarios with varying gaps between vehicles. First results indicate an acceptable level of realism so that the device can be used for further user studies addressing pedestrian behavior, pedestrian interaction with (automated) vehicles, risk assessment and investigation of the pre-crash phase without the risk of injuries.

A Novel Approach for the Test of Active Pedestrian Safety Systems

Active pedestrian safety systems are based on a variety of sensor systems and detection algorithm approaches. The activation of, e.g., an emergency brake is a critical decision. Therefore, such applications must be tested very responsibly. This paper shows the characteristic features of pedestrians, presents current test methods, and introduces a novel test system approach and pedestrian dummy, which enables the test of advanced pedestrian detection systems. This includes also detection algorithm approaches including pre-indicators and path prediction for a complex motion pattern. The complex motion pattern can be a walking or a running pedestrian, including velocity and direction changes.

Dynamic position calibration by road structure detection

In this research study, the potential of landmarks for accuracy improvement in terms of satellite-based localization is analyzed. Advanced driver assistance systems (ADAS) and vehicle-to-vehicle (V2V) communication applications benefit from precise determination of position. Especially applications for cooperative car safety work best when the position determination is as accurate as possible. The novel approach is related to the differential GPS method, but without using a base station. Previously measured landmarks are used instead. To determine the potential of the approach, the level of increasing localization accuracy is evaluated. Moreover the distribution of naturally existing landmarks is analyzed using the recorded data of an endurance testing vehicle.

Test Methodology for Automotive Surround Sensors in Dynamic Driving Situations

Modern vehicles use surround sensors to measure their local environment. The information are processed and forwarded to intelligent pre-crash or automation functions enhancing vehicle safety or enabling automated driving. False or inaccurate measurements can lead to fatal consequences for humans and vehicles. High accuracy and robust environment perception in all driving situations, including high dynamic driving situations e.g. skidding, are compulsory requirements for these systems. Therefore, automotive surround sensors must be tested in various driving situations. This paper presents a new, non-destructive and reproducible test methodology for testing surround sensors in high dynamic driving situations. Therefore, the vehicles motion during a skid driving situation was mathematically described. The test methodology was validated through experiments carried out with a real test vehicle. Finally, the experimental setup and the results are presented and discussed.

Reproducible Fog Simulation for Testing Automotive Surround Sensors

Intelligent vehicles use surround sensors such as radar, lidar and camera to perceive their local environment. The generated information are processed by a control unit to identify critical traffic situations such as rear-end collisions and trigger reversible or irreversible safety systems. Incorrect measurements can result in accidents with fatal consequences. Therefore, high reliability and accuracy is a mandatory requirement. The performance of surround sensors depend on the ambient atmosphere and weather conditions. It is known that fog has a negative influence on wave propagation especially in the visible range. Hence, surround sensors must be tested under realistic conditions. To cope with this problem, this paper presents a novel setup together with a test methodology to investigate the influence of fog on automotive radar, lidar and camera sensors. Based on static indoor tests, we have shown that the measurements are consistent with literature and enable reproducible testing.

Potential of Plenoptic Cameras in the Field of Automotive Safety

Cameras in general are exteroceptive sensors and provide visual information about the local environment. For automotive safety applications, image data is processed to identify critical situations and avoid traffic accidents. Plenoptic cameras which are capable of recording both, angular and spatial information of light, have found increasing interest in optical 3D measurement techniques in recent years. Major advantages are multiple perspective views of captured scene and post focus capability. Therefore, the depth of the scene can be estimated similar to stereo camera.