Abstract:Radar sensors were among the first perceptual sensors used for automated driving. Although several other technologies such as lidar, camera, and ultrasonic sensors are available, radar sensors have maintained and will continue to maintain their importance due to their reliability in adverse weather conditions. Virtual methods are being developed for verification and validation of automated driving functions to reduce the time and cost of testing. Due to the complexity of modelling high-frequency wave propagati… Show more
“…As can be seen, the requirement for simulation efficiency in sensor modelling was met, as neither increased computing power nor special code optimisation was required to run the simulation faster than real time for all three in Section 1 . For more details about sensor classification, please refer to [ 2 ].…”
Section: Resultsmentioning
confidence: 99%
“…To support X-in-the-loop testing methods during the vehicle development process, a wide range of commercial or open-source simulation software is available to the automotive industry. Referring to our previous work [ 2 ], some examples are given: in [ 5 ]: TASS-PreScan, dSpace-ASM, in [ 3 ]: TESIS Dyna4-Driver Assistance, MathWorks-ADAS Toolbox, in [ 6 ]: CARLA, AirSim, DeepDrive, Udacity, or in [ 7 ]: CarMaker from IPG Automotive GmbH., VIRES-VTD. These software packages provide a variety of interfaces for modelling perception sensors at different levels of complexity, but their parallel use is often limited, whereas in real application data, the fusion of multiple sensors is state of the art.…”
Section: Related Workmentioning
confidence: 99%
“…Due to the above-mentioned need for virtual V&V, sensor models of different complexity and accuracy are required at different stages of vehicle development to fully satisfy the respective requirements. Therefore, in a previous work [ 2 ], we introduced a novel method for classifying radar sensor models found in the literature according to the stages of the vehicle development process. Based on the vehicle manufacturers’ requirements, we have also assigned them to these stages depending on their applicability.…”
Despite the progress in driving automation, the market introduction of higher-level automation has not yet been achieved. One of the main reasons for this is the effort in safety validation to prove functional safety to the customer. However, virtual testing may compromise this challenge, but the modelling of machine perception and proving its validity has not been solved completely. The present research focuses on a novel modelling approach for automotive radar sensors. Due to the complex high-frequency physics of radars, sensor models for vehicle development are challenging. The presented approach employs a semi-physical modelling approach based on experiments. The selected commercial automotive radar was applied in on-road tests where the ground truth was recorded with a precise measurement system installed in ego and target vehicles. High-frequency phenomena were observed and reproduced in the model on the one hand by using physically based equations such as antenna characteristics and the radar equation. On the other hand, high-frequency effects were statistically modelled using adequate error models derived from the measurements. The model was evaluated with performance metrics developed in previous works and compared to a commercial radar sensor model. Results show that, while keeping real-time performance necessary for X-in-the-loop applications, the model is able to achieve a remarkable fidelity as assessed by probability density functions of the radar point clouds and using the Jensen–Shannon divergence. The model delivers values of radar cross-section for the radar point clouds that correlate well with measurements comparable with the Euro NCAP Global Vehicle Target Validation process. The model outperforms a comparable commercial sensor model.
“…As can be seen, the requirement for simulation efficiency in sensor modelling was met, as neither increased computing power nor special code optimisation was required to run the simulation faster than real time for all three in Section 1 . For more details about sensor classification, please refer to [ 2 ].…”
Section: Resultsmentioning
confidence: 99%
“…To support X-in-the-loop testing methods during the vehicle development process, a wide range of commercial or open-source simulation software is available to the automotive industry. Referring to our previous work [ 2 ], some examples are given: in [ 5 ]: TASS-PreScan, dSpace-ASM, in [ 3 ]: TESIS Dyna4-Driver Assistance, MathWorks-ADAS Toolbox, in [ 6 ]: CARLA, AirSim, DeepDrive, Udacity, or in [ 7 ]: CarMaker from IPG Automotive GmbH., VIRES-VTD. These software packages provide a variety of interfaces for modelling perception sensors at different levels of complexity, but their parallel use is often limited, whereas in real application data, the fusion of multiple sensors is state of the art.…”
Section: Related Workmentioning
confidence: 99%
“…Due to the above-mentioned need for virtual V&V, sensor models of different complexity and accuracy are required at different stages of vehicle development to fully satisfy the respective requirements. Therefore, in a previous work [ 2 ], we introduced a novel method for classifying radar sensor models found in the literature according to the stages of the vehicle development process. Based on the vehicle manufacturers’ requirements, we have also assigned them to these stages depending on their applicability.…”
Despite the progress in driving automation, the market introduction of higher-level automation has not yet been achieved. One of the main reasons for this is the effort in safety validation to prove functional safety to the customer. However, virtual testing may compromise this challenge, but the modelling of machine perception and proving its validity has not been solved completely. The present research focuses on a novel modelling approach for automotive radar sensors. Due to the complex high-frequency physics of radars, sensor models for vehicle development are challenging. The presented approach employs a semi-physical modelling approach based on experiments. The selected commercial automotive radar was applied in on-road tests where the ground truth was recorded with a precise measurement system installed in ego and target vehicles. High-frequency phenomena were observed and reproduced in the model on the one hand by using physically based equations such as antenna characteristics and the radar equation. On the other hand, high-frequency effects were statistically modelled using adequate error models derived from the measurements. The model was evaluated with performance metrics developed in previous works and compared to a commercial radar sensor model. Results show that, while keeping real-time performance necessary for X-in-the-loop applications, the model is able to achieve a remarkable fidelity as assessed by probability density functions of the radar point clouds and using the Jensen–Shannon divergence. The model delivers values of radar cross-section for the radar point clouds that correlate well with measurements comparable with the Euro NCAP Global Vehicle Target Validation process. The model outperforms a comparable commercial sensor model.
“…So far, hardly any ontologies for autonomous driving have been found, which focus on the detailed, physical level. However, when extending the search to sensor-based testing, many different approaches are discussed, especially in the radar domain [44].…”
The verification and validation of autonomous driving vehicles remains a major challenge due to the high complexity of autonomous driving functions. Scenario-based testing is a promising method for validating such a complex system. Ontologies can be utilized to produce test scenarios that are both meaningful and relevant. One crucial aspect of this process is selecting the appropriate method for describing the entities involved. The level of detail and specific entity classes required will vary depending on the system being tested. It is important to choose an ontology that properly reflects these needs.This paper summarizes key representative ontologies for scenario-based testing and related use cases in the field of autonomous driving. The considered ontologies are classified according to their level of detail for both static facts and dynamic aspects. Furthermore, the ontologies are evaluated based on the presence of important entity classes and the relations between them.
“…2 of 27 detections are far too sparse [13,14]. The sparse nature of radar point clouds collected with many vehicular radars (usually 64 points or less) might explain this [15].…”
AVs suffer reduced maneuverability and performance due to the degradation in sensor performances in fog. Such degradation causes significant object detection errors essential for AVs' safety-critical conditions. For instance, YOLOv5 performs significantly well under favorable weather but suffers miss detections and false positives due to atmospheric scattering caused by fog particles. Existing deep object detection techniques often exhibit a high degree of accuracy. The drawback is being sluggish at object detection in fog. Object detection methods with fast detection speed have been obtained using deep learning at the expense of accuracy. The problem of the lack of balance between detection speed and accuracy in fog persist. This paper presents an improved YOLOv5-based multi-sensor fusion network that combines radar's object detection with a camera image bounding box. We transformed radar detection by mapping the radar detections into a two-dimensional image coordinate and projected the resultant radar image on the camera image. Using the attention mechanism, we emphasized and improved important feature representation used for object detection while reducing high-level feature information loss. We trained and tested our multi-sensor fusion network on clear and multi-fog weather datasets obtained from the CARLA simulator. Our result shows that the proposed method significantly enhances the detection of distant and small objects. Our small CR-YOLOnet model best strikes a balance between accuracy and speed with an accuracy of 0.849 at 69 fps.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.