| 1. |
- Fredriksson, Rikard, 1967, et al.
(författare)
-
Integrated bicyclist protection systems - potential of head injury reduction combining passive and active protection systems
- 2015
-
Ingår i: 24th International Technical Conference on the Enhanced Safety of Vehicles.
-
Konferensbidrag (refereegranskat)abstract
- In recent years both pedestrian passive and active safety systems, such as pedestrian bonnets/airbags and autonomous braking, have emerged on the market and are estimated to be effective to reduce injury of vulnerable road users in car crashes. A natural next step is to develop similar protection systems for bicyclists. The aim of this study was to investigate the potential bicyclist head injury reduction from passive and active protection systems compared to an integrated system. The German In-Depth Accident Study (GIDAS) database was queried from 1999 to 2014 for severely (AIS3+) head injured bicyclists when struck by passenger car fronts. This resulted in 34 cases where information was sufficient for both the pre-crash and the in-crash part of the event. The default passive protection system was designed to mitigate head injuries caused by the bonnet area, A-pillars, and the lower windscreen (instrument panel) area (deployable hood and windshield airbag). To estimate the hood and airbag performance risk reduction functions were used based on experimental tests with and without the systems. The active protection system was an autonomous braking system, which was activated one second prior to impact if the bicyclist was visible to a forward-looking sensor. Maximum speed reduction was estimated using road condition information in each case. The integrated system was a direct combination of the passive and active protection systems. Case by case the effect from each of the active, passive and integrated systems was estimated. For the integrated system, the influence of the active system on the passive system performance was explicitly modelled in each case. A sensitivity analysis was performed varying the coverage area of the passive protection system and the activation criteria of the active system. The integrated system resulted in 29%-62% higher effectiveness than the best single system of active respectively passive protection system in reducing the number of bicyclists sustaining severe (AIS3+) head injuries. These values were statistically tested and found to be significant. The study is based on representative data from Germany, but may not be representative to countries with a different car fleet or infrastructure. This study indicates that integrated systems of passive and active vulnerable road user countermeasures offer a significantly increased potential for head injury reduction compared to either of the two systems alone.
|
|
| 2. |
- Hornauer, Sascha, et al.
(författare)
-
Driving scene retrieval by example from large-scale data
- 2019
-
Ingår i: CVPR Workshops 2019.
-
Konferensbidrag (refereegranskat)abstract
- Many machine learning approaches train networks with input from large datasets to reach high task performance. Collected datasets, such as Berkeley Deep Drive Video (BDD-V) for autonomous driving, contain a large variety of scenes and hence features. However, depending on the task, subsets, containing certain features more densely, support training better than others. For example, training networks on tasks such as image segmentation, bounding box detection or tracking requires an ample amount of objects in the input data. When training a network to perform optical flow estimation from first-person video, over-proportionally many straight driving scenes in the training data may lower generalization to turns. Even though some scenes of the BDD-V dataset are labeled with scene, weather or time of day information, these may be too coarse to filter the dataset best for a particular training task. Furthermore, even defining an exhaustive list of good label-types is complicated as it requires choosing the most relevant concepts of the natural world for a task. Alternatively, we investigate how to use examples of desired data to retrieve more similar data from a large-scale dataset. Following the paradigm of ”I know it when I see it”, we present a deep learning approach to use driving examples for retrieving similar scenes from the BDD-V dataset. Our method leverages only automatically collected labels. We show how we can reliably vary time of the day or objects in our query examples and retrieve nearest neighbors from the dataset. Using this method, already collected data can be filtered to remove bias from a dataset, removing scenes regarded too redundant to train on.
|
|
| 3. |
- Lübbe, Nils, 1982, et al.
(författare)
-
Predicted road traffic fatalities in Germany: The potential and limitations of vehicle safety technologies from passive safety to highly automated driving
- 2018
-
Ingår i: Conference proceedings International Research Council on the Biomechanics of Injury, IRCOBI. - 2235-3151. ; 2018-September, s. 17-52
-
Konferensbidrag (refereegranskat)abstract
- It has been proposed that automated vehicles will greatly increase road traffic safety. However, few attempts have been made to quantify this thesis and to compare the expected benefits with more traditional safety systems. This study was carried out in five steps, adding systems in each step (from passive safety, standard Advances Driver Assistance Systems (ADAS), advanced ADAS, safety-minded driving, to cautious driving) in order to capture the benefit of increasing levels of automation. Conservative and optimistic rules based on the expected performance of each safety system were developed and applied to the German In-Depth Accident Study database. Adding safety systems was effective in preventing fatalities, ranging from 12-13% (step 1, passive safety, no automation, conservative-optimistic estimate) to 45-63% (step 5, cautious driving). The highest automation level, in step 5, achieved a reduction of Vulnerable Road User (VRU) fatalities of 33-41%. Thus, passive and active safety systems contribute substantially to preventing fatalities and their further development and deployment should not be abandoned. Even the safest foreseeable, highly automated passenger cars are not likely to avoid all crashes and all road traffic fatalities. While increased market penetration across safety systems will make road traffic substantially safer, more efforts are needed to protect VRUs.
|
|
| 4. |
- op den Camp, Olaf, et al.
(författare)
-
Overview of main accident scenarios in car-tocyclist accidents for use in AEB-system test protocol
- 2014
-
Konferensbidrag (övrigt vetenskapligt/konstnärligt)abstract
- The overall number of fatalities in road traffic accidents in Europe is decreasing. Unfortunately, the number of fatalities among cyclists does not follow this trend with the same rate [5]. In the Netherlands, a major share of killed cyclists in traffic accidents was the result of a collision with a motorised vehicle [2]. The automotive industry is making a significant effort in the development and implementation of safety systems in cars to avoid or mitigate an imminent crash with vulnerable road users, and more specifically with cyclists. The current state‐of‐the‐art of active safety systems, Autonomous Emergency Braking (AEB), is being widely introduced. A car equipped with AEB makes use of onboard sensors such as camera and radar, to track and trace traffic participants that possibly interfere with the trajectory of the car. This information is used to warn the driver in case of a possibly critical situation and/or to brake in case the driver does not respond and the risk of collision does not decrease. Currently, AEB systems that are designed to avoid car‐to‐car collisions are part of the Euro NCAP star rating. In 2016, Euro NCAP will include AEB systems for pedestrians in the star rating. It is the intention of Euro NCAP to include AEB systems for cyclists in the star rating beginning of 2018 [3]. To support and prepare the introduction of Cyclist‐AEB systems and the resulting consumer tests of such systems, TNO has taken the initiative to set‐up a consortium of car manufacturers and suppliers with the support of Euro NCAP laboratories (such as BASt) to develop a testing system and test protocol for Cyclist‐AEB systems. This paper reports the first steps towards this protocol in which an indepth road accident study is performed to determine what accident scenarios are most relevant for car‐to‐cyclist collisions. Data of killed and seriously injured cyclists due to collision with a passenger car were included in this study. An overview is given for the following European countries: Germany, the Netherlands, Sweden, France, Italy, and the United Kingdom.
|
|
| 5. |
- Ranjbar, Arian, 1992, et al.
(författare)
-
Safety Monitoring of Neural Networks Using Unsupervised Feature Learning and Novelty Estimation
- 2022
-
Ingår i: IEEE Transactions on Intelligent Vehicles. - 2379-8858. ; 7:3, s. 711-721
-
Tidskriftsartikel (refereegranskat)abstract
- Neural networks are currently suggested to be implemented in several different driving functions of autonomous vehicles. While showing promising results the drawback lies in the difficulty of safety verification and ensuring operation as intended. The aim of this paper is to increase safety when using neural networks, by proposing a monitoring framework based on novelty estimation of incoming driving data. The idea is to use unsupervised instance discrimination to learn a similarity measure across ego-vehicle camera images. By estimating a von Mises-Fisher distribution of expected ego-camera images they can be compared with unexpected novel images. A novelty measurement is inferred through the likelihood of test frames belonging to the expected distribution. The suggested method provides competitive results to several other novelty or anomaly detection algorithms on the CIFAR-10 and CIFAR-100 datasets. It also shows promising results on real world driving scenarios by distinguishing novel driving scenes from the training data of BDD100k. Applied on the identical training-test data split, the method is also able to predict the performance profile of a segmentation network. Finally, examples are provided on how this method can be extended to find novel segments in images.
|
|
| 6. |
- Ranjbar, Arian, 1992, et al.
(författare)
-
Scene Novelty Prediction from Unsupervised Discriminative Feature Learning
- 2020
-
Ingår i: 2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC).
-
Konferensbidrag (refereegranskat)abstract
- Deep learning approaches are widely explored in safety-critical autonomous driving systems on various tasks. Network models, trained on big data, map input to probable prediction results. However, it is unclear how to get a measure of confidence on this prediction at the test time. Our approach to gain this additional information is to estimate how similar test data is to the training data that the model was trained on. We map training instances onto a feature space that is the most discriminative among them. We then model the entire training set as a Gaussian distribution in that feature space. The novelty of the test data is characterized by its low probability of being in that distribution, or equivalently a large Mahalanobis distance in the feature space. Our distance metric in the discriminative feature space achieves a better novelty prediction performance than the state-of-the-art methods on most classes in CIFAR-10 and ImageNet. Using semantic segmentation as a proxy task often needed for autonomous driving, we show that our unsupervised novelty prediction correlates with the performance of a segmentation network trained on full pixel-wise annotations. These experimental results demonstrate potential applications of our method upon identifying scene familiarity and quantifying the confidence in autonomous driving actions.
|
|
| 7. |
- Ranjbar, Arian, 1992
(författare)
-
Towards Safe Autonomous Driving
- 2021
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Autonomous driving is expected to bring several benefits, in particular regarding safety. This thesis aim to contribute towards two questions concerning safety: "What is the potential safety benefit of autonomous driving?'' and "How can we ensure safe operation of such vehicles?''. In the first part of the thesis, methods for evaluating the safety benefit are investigated. In particular predictive effectiveness evaluation based on resimulation of accident data, using models to estimate new outcomes in case the safety system had been available. To illustrate the methodology, four examples of gradual increase in model complexity are presented. First, an Autonomous Emergency Braking (AEB) system using a sensor model, decision algorithm, vehicle dynamics model and regression based injury model. This is extended in a Forward Collision Warning (FCW) system which additionally requires a driver model to simulate driver reactions. The third example shows how an active, AEB, and passive, airbag, system can be combined. Finally the fourth example combines several systems to emulate a highly automated vehicle. Apart from predicting the real world performance, this analysis also identifies current safety gaps by studying the residual of the accident set. Safety benefit estimation using accident data gives an evaluation on the current accident distributions, however, the systems may introduce new accidents if not operated as intended. In the second part of the thesis, safety verification processes with the intent of preventing unsafe operation, are presented. This is particularly challenging for machine learning based components, such as neural networks. In this case, traditional analytical verification approaches are difficult to apply due to the non-linearity and high dimensional parameter spaces. Similarly, statistical safety arguments often require unfeasible amounts of annotated validation data. Instead, monitor functions are investigated as a complement to increase safety during operation. The method presented estimates the similarity of the driving environment, compared to the training data, where decisions inferred from novel data can be considered less reliable. Although not providing a complete safety assurance, the methodology show promising initial results for increasing safety. In addition, it could potentially be used to collect novel data and reduce redundancy in training data.
|
|
