Search

This paper proposes a deep learning based object detection method to locate a distant region in an image in real-time. It concentrates on distant objects from a vehicular front camcorder perspective, trying to solve one of the common problems in Advanced Driver Assistance Systems (ADAS) applications, which is, to detect the smaller and faraway objects with the same confidence as those with the bigger and closer objects. This paper presents an efficient multi-scale object detection network, termed as ConcentrateNet to detect a vanishing point and concentrate on the near-distant region. Initially, the object detection model inferencing will produce a larger scale of receptive field detection results and predict a potentially vanishing point location, that is, the farthest location in the frame. Then, the image is cropped near the vanishing point location and processed with the object detection model for second inferencing to obtain distant object detection results. Finally, the two-inferencing results are merged with a specific Non-Maximum Suppression (NMS) method. The proposed network architecture can be employed in most of the object detection models as the proposed model is implemented in some of the state-of-the-art object detection models to check feasibility. Compared with original models using higher resolution input size, ConcentrateNet architecture models use lower resolution input size, with less model complexity, achieving significant precision and recall improvements. Moreover, the proposed ConcentrateNet architecture model is successfully ported onto a low-powered embedded system, NVIDIA Jetson AGX Xavier, suiting the real-time autonomous machines.

Collision avoidance systems are of great importance in advanced driver assistance systems (ADAS) to increase the level of safety. However, as a result of their excessive costs, are not considered as a feature in lower-end vehicles. The concept of vehicular communication has emerged recently and added a new dimension for promoting safety in automotive. In this paper, the opportunities for implementation of the cost-effective ADAS through next generation of vehicular communication are discussed. A conceptual model for a vision-based driver assistance system based on collaborative mobile edge computing (MEC) is proposed. Cloudlets are the edges of the 5G cellular network and are capable of providing MEC solutions to their connected devices. The proposed system monitors the vehicles driving in front of an ego car and notifies the driver upon detecting hazardous conditions caused by other vehicles movements. This system can be employed as a cost-effective driver assistance system and has the potential to reach the mass market and be used in all ranges of vehicles including the lower-end models. [ABSTRACT FROM AUTHOR]

The Mobile Millennium project provided a platform for aggregating traffic information across various sources, including infrastructure sensors, commercial data feeds, probe vehicles, and probe cell phones. The Networked Traveler project provided the California PATH instrumented research vehicle platform used to both deliver vehicle probe data back to the infrastructure and to generate Advanced Driver Assistance Systems (ADAS) alerts to the drivers of those vehicles. The main theme of this collaboration project was to demonstrate the potential to create Enhanced Probe Vehicles (EPVs) by merging vehicle CAN-bus data with the typical GPS data that is provided by normal probe vehicles. The first goal of this project was integrate the two systems, allowing the instrumented research vehicles to communicate with the Mobile Millennium traffic aggregation servers. The second goal was to enhance the data gathered and stored by the Mobile Millennium traffic aggregation server to include the enhanced information provided by the EPVs, specifically, to include hazard warning light activation events. The third and final goal of this project was to integrate the information provided by the EPVs to the traffic server into soft-safety alerts that could be provided to the drivers of the EPVs and demonstrate the prototype system. This report documents the systems that were built for the demonstration, and provides a short analysis of the system performance from data gathered through both a simulation and an on-the-road vehicle test

Compared with the lane keeping maneuvers, lane changing maneuvers are much more complicated. Inappropriate ones may result in traffic accidents. Therefore, it is necessary to develop vehicular social networks (VSNs) and advanced driver assistance systems (ADAS), which can help to evaluate the traffic situation and provide additional warnings for drivers' unsafe lane changing operations. This paper proposes a deep learning model to simulate the situation assessment and decision making process during lane changing events. Compared with the existing assessment models, our model has two significant advantages. First, except for the instantaneous states of the subject and the surrounding vehicles, the proposed model also takes drivers' historical experience and the vehicle-to-vehicle (V2V) memory effect into consideration for the final lane changing maneuvers situation assessment. Second, our new assessment model is built based on Deep Neural Networks (DNNs), which is proved to outperform conventional machine learning classifiers. The empirical trajectory dataset NGSIM is used for evaluating the performance of the proposed model. Experiment results indicate that our model achieves high identification accuracy of both lane changing maneuvers and lane keeping maneuvers. It is verified to be an effective driverless technology in assisting vehicle warning systems.

Pedestrian detection is one of the important computer vision problems in automotive safety and driver assistance domain. It is a major component of the advanced driver assistance system (ADAS) which help the driver to drive safely. Recent literature shows a number of research activities addressing object detection/tracking in general and pedestrian detection in particular. The solutions proposed by different researchers vary in detection methods, detection scenario, feature descriptors, classification schemes, detection performance, as well as computational complexity. However, the average detection accuracy is not much promising even after many years of research. The fail-safe and real-time human detection from real life road scenes, even in standard resolution, is far from reality. Safety critical systems in the automotive industry have to follow well established stringent safety standards like ISO26262. Since the pedestrian detection system deals with human safety, it also has to follow these standards before integrating to the vehicle electronics. This paper is a study of different techniques used in pedestrian detection specific to the automotive application, along with a description of generic pedestrian detection solution architecture.

Advanced Driver Assistance Systems (ADAS), created for enhancing the driving experience and actively preventing road crashes, have been progressively incorporated in vehicle designing essentially during the last decade. However, the literature has shown how some of these assisting technologies are not used by drivers in tandem with their potential. The aims of this study were, first, to examine the availability and demand of ADAS technologies among Spanish drivers and, secondly, to explore the perceived constraints and discouraging reasons for avoiding the use of ADAS available in their vehicles. For this national cross-sectional study, data from 1,207 Spanish drivers were analyzed. The results of this study show that, on one hand, GPS navigation, rain sensors and automatic lighting are the most frequently used ADAS features in Spain and, on the other, that gestural control, E-call and post-collision emergency braking are the less demanded. Also, there are age and gender-based differences in the valuation of certain ADAS features. Further, low perceived value, lack of confidence and potential distractibility constitute the main constraints perceived by drivers to actively use these assisting technologies while driving. In this regard, and jointly with a progressive vehicle automatization, a deeper emphasis on driver training, safety and efficiency-related benefits of ADAS technologies may strengthen its acceptance and progressive inclusion in everyday driving.

Assistive vision-based solutions for the driver extend the capabilities of human vision and support safe travel. Unfortunately, their widespread usage is generally limited to expensive cars. Interestingly, a high price is most likely a derivative of the costs incurred in the research instead of the value of hardware components. In the article we show that a mobile system for pedestrian detection in severe lighting conditions can be build using state of the art algorithms and widely available hardware. The proposed night-vision system for pedestrian detection processes thermal images using a proprietary ODROID XU4 microcomputer under Ubuntu MATE operating system. We applied a cascade object detector for the task of human silhouette detection in context of thermal imagery and contrasted the results with the state of the art deep learning approach. The experiments conducted prove effectiveness of the proposed solution. [ABSTRACT FROM AUTHOR]

Since 2017, a research team from the Technical University of Munich has developed a software stack for autonomous driving. The software was used to participate in the Roborace Season Alpha Championship. The championship aims to achieve autonomous race cars competing with different software stacks against each other. In May 2019, during a software test in Modena, Italy, the greatest danger in autonomous driving became reality: A minor change in environmental influences led an extensively tested software to crash into a barrier at speed. Crashes with autonomous vehicles have happened before but a detailed explanation of why software failed and what part of the software was not working correctly is missing in research articles. In this paper we present a general method that can be used to display an autonomous vehicle disengagement to explain in detail what happened. This method is then used to display and explain the crash from Modena. Firstly a brief introduction into the modular software stack that was used in the Modena event, consisting of three individual parts—perception, planning, and control—is given. Furthermore, the circumstances causing the crash are elaborated in detail.By presented and explaining in detail which software part failed and contributed to the crash we can discuss further software improvements. As a result, we present necessary functions that need to be integrated in an autonomous driving software stack to prevent such a vehicle behavior causing a fatal crash. In addition we suggest an enhancement of the current disengagement reports for autonomous driving regarding a detailed explanation of the software part that was causing the disengagement. In the outlook of this paper we present two additional software functions for assessing the tire and control performance of the vehicle to enhance the autonomous. [ABSTRACT FROM AUTHOR]

This paper studies a control architecture for vehicle lateral dynamics based on the execution of optimal trajectories via feedforward inverse model control. The focus here is on assessing the robustness of this arrangement when the vehicle real forward dynamics is not identically cancelled by the inverse model due to model approximations and parameter uncertainties. The trajectories that are considered are analytic solutions of the minimum square jerk optimal control problem for a simplified kinematic vehicle model in curvilinear coordinates. Various hypotheses are made concerning the mismatch between the inverse model and the actual forward dynamics of the vehicle. Closed-loop stability analysis shows that the studied control scheme guarantees asymptotic stability of both the nominal and the perturbed kinematic model with significant robustness margins. In addition to the theoretical robustness analysis, the same control scheme is validated in simulation using the industry-standard software for virtual vehicle testing IPG CarMaker. [ABSTRACT FROM AUTHOR]

Abstract This paper presents a robust method for generic obstacle detection and collision warning in Advanced Driver Assistance Systems (ADAS). The highlight of our method is the ability to detect all obstacles without prior knowledge and detect partially occluded obstacles including the obstacles that have not completely appeared in the frame (truncated obstacles). Our results show an improvement of 90% more true positives per frame compared to one of the state-of-the-art methods. Our proposed method is robust to variations in illumination and to a wide variety of vehicles and obstacles that are encountered while driving. Distortions are introduced when Inverse Perspective Mapping (IPM) projects the camera image onto the road surface plane. In this paper, we first show that the angular distortion in the IPM domain belonging to obstacle edges varies as a function of their corresponding 2D location in the camera plane. We use this information to perform proposal generation. We propose a novel proposal assessment method based on fusing statistical properties from both the IPM image and the camera image to perform robust outlier elimination and false positive reduction. We also present an annotated obstacle detection dataset derived from various source videos that can serve as a benchmark for the evaluation of future obstacle detection algorithms. The source videos containing diverse illumination and traffic conditions are derived from multiple publicly available datasets. Highlights • Robust method for generic obstacle detection and collision warning in Advanced Driver Assistance Systems (ADAS). • Ability to detect all obstacles without prior knowledge and detect partially occluded and truncated obstacles. • Results can produce up to 90% more true positives per frame compared to the existing methods on the tested videos. • Proposed method is shown to provide accurate detection in various illumination conditions such as daylight, twilight, night and rainy conditions. [ABSTRACT FROM AUTHOR]

This paper presents a human-centered risk assessment algorithm using vehicular communication for application to an automated driving vehicle. Vehicle-to-vehicle (V2V) wireless communication has been implemented and fused with a radar sensor to obtain the prediction of the remote vehicle’s motion. Based on the predicted behavior of remote vehicles, a collision risk and a human reaction time are determined for a human-centered active safety control intervention moment. The human-centered risk assessment algorithm has been incorporated into a collision avoidance algorithm to monitor threat vehicles ahead and to find the best intervention point. Effects of the vehicular communication on a perception and a control performance are investigated. The performance of the proposed algorithm has been investigated via computer simulations and vehicle tests. It has been shown from both simulations and vehicle tests that the proposed human-centered risk assessment algorithm with V2V communication decides a proper active safety intervention moment by reducing the chances of over/underestimate of a conventional radar-only system. [ABSTRACT FROM AUTHOR]

This paper proposes a real-time detection method for a car driving ahead in real time on a tunnel road. Unlike the general road environment, the tunnel environment is irregular and has significantly lower illumination, including tunnel lighting and light reflected from driving vehicles. The environmental restrictions are large owing to pollution by vehicle exhaust gas. In the proposed method, a real-time detection method is used for vehicles in tunnel images learned in advance using deep learning techniques. To detect the vehicle region in the tunnel environment, brightness smoothing and noise removal processes are carried out. The vehicle region is learned after generating a learning image using the ground-truth method. The YOLO v2 model, with an optimal performance compared to the performances of deep learning algorithms, is applied. The training parameters are refined through experiments. The vehicle detection rate is approximately 87%, while the detection accuracy is approximately 94% for the proposed method applied to various tunnel road environments.

Test, verification, and development activities of vehicles with ADAS (Advanced Driver Assistance Systems) and ADF (Automated Driving Functions) generate large amounts of measurement data. To efficiently evaluate and use this data, a generic understanding and classification of the relevant driving scenarios is necessary. Currently, such understanding is obtained by using heuristic algorithms or even by manual inspection of sensor signals. In this paper, we apply deep learning on sensor time series data to automatically extract relevant features for classification of driving scenarios relevant for a Lane-Keep-Assist System. We compare the performance of convolutional and recurrent neural networks and propose two classification models. The first one is an online model for scenario classification during driving. The second one is an offline model for post-processing, providing higher accuracy. [ABSTRACT FROM AUTHOR]

Intelligent connected vehicles (ICVs) are believed to change people’s life in the near future by making the transportation safer, cleaner and more comfortable. Although many prototypes of ICVs have been developed to prove the concept of autonomous driving and the feasibility of improving traffic efficiency, there still exists a significant gap before achieving mass production of high-level ICVs. The objective of this study is to present an overview of both the state of the art and future perspectives of key technologies that are needed for future ICVs. It is a challenging task to review all related works and predict their future perspectives, especially for such a complex and interdisciplinary area of research. This article is organized to overview the ICV key technologies by answering three questions: what are the milestones in the history of ICVs; what are the electronic components needed for building an ICV platform; and what are the essential algorithms to enable intelligent driving? To answer the first question, the article has reviewed the history and the development milestones of ICVs. For the second question, the recent technology advances in electrical/electronic architecture, sensors, and actuators are presented. For the third question, the article focuses on the algorithms in decision making, as the perception and control algorithm are covered in the development of sensors and actuators. To achieve correct decision-making, there exist two different approaches: the principle-based approach and data-driven approach. The advantages and limitations of both approaches are explained and analyzed. Currently automotive engineers are concerned more with the vehicle platform technology, whereas the academic researchers prefer to focus on theoretical algorithms. However, only by incorporating elements from both worlds can we accelerate the production of high-level ICVs. [ABSTRACT FROM AUTHOR]

High-end vehicles are already equipped with safety systems, such as assistive braking and automatic lane following, enhancing vehicle safety. Yet, these current solutions can only help in low-complexity driving situations. In this paper, we introduce a parallel autonomy, or shared control, framework that computes safe trajectories for an automated vehicle, based on human inputs. We minimize the deviation from the human inputs while ensuring safety via a set of collision avoidance constraints. Our method achieves safe motion even in complex driving scenarios, such as those commonly encountered in an urban setting. We introduce a receding horizon planner formulated as nonlinear model predictive control (NMPC), which includes the analytic descriptions of road boundaries and the configuration and future uncertainties of other road participants. The NMPC operates over both steering and acceleration simultaneously. We introduce a nonslip model suitable for handling complex environments with dynamic obstacles, and a nonlinear combined slip vehicle model including normal load transfer capable of handling static environments. We validate the proposed approach in two complex driving scenarios. First, in an urban environment that includes a left-turn across traffic and passing on a busy street. And second, under snow conditions on a race track with sharp turns and under complex dynamic constraints. We evaluate the performance of the method with various human driving styles. We consequently observe that the method successfully avoids collisions and generates motions with minimal intervention for parallel autonomy. We note that the method can also be applied to generate safe motion for fully autonomous vehicles. [ABSTRACT FROM AUTHOR]

Typically, lane departure warning systems rely on lane lines being present on the road.However, in many scenarios, e.g., secondary roads or some streets in cities, lane lines are eithernot present or not sufficiently well signaled. In this work, we present a vision-based method tolocate a vehicle within the road when no lane lines are present using only RGB images as input.To this end, we propose to fuse together the outputs of a semantic segmentation and a monoculardepth estimation architecture to reconstruct locally a semantic 3D point cloud of the viewed scene.We only retain points belonging to the road and, additionally, to any kind of fences or walls thatmight be present right at the sides of the road. We then compute the width of the road at a certainpoint on the planned trajectory and, additionally, what we denote as the fence-to-fence distance.Our system is suited to any kind of motoring scenario and is especially useful when lane lines arenot present on the road or do not signal the path correctly. The additional fence-to-fence distancecomputation is complementary to the road’s width estimation. We quantitatively test our methodon a set of images featuring streets of the city of Munich that contain a road-fence structure, so asto compare our two proposed variants, namely the road’s width and the fence-to-fence distancecomputation. In addition, we also validate our system qualitatively on the Stuttgart sequence of thepublicly available Cityscapes dataset, where no fences or walls are present at the sides of the road,thus demonstrating that our system can be deployed in a standard city-like environment. For thebenefit of the community, we make our software open source.

Driving safety field (DSF) model has been proposed to represent comprehensive driving risk formed by interactions of driver-vehicle-road in mixed traffic environment. In this work, we establish an optimization model based on grey relation degree analysis to calibrate risk coefficients of DSF model. To solve the optimum solution, a genetic algorithm is employed. Finally, the DSF model is verified through a real-world driving experiment. Results show that the DSF model is consistent with driver's hazard perception and more sensitive than TTC. Moreover, the proposed DSF model offers a novel way for criticality assessment and decision-making of advanced driver assistance systems and intelligent connected vehicles. [ABSTRACT FROM AUTHOR]

Test, verification, and development activities of vehicles with ADAS (Advanced Driver Assistance Systems) and ADF (Automated Driving Functions) generate large amounts of measurement data. To efficiently evaluate and use this data, a generic understanding and classification of the relevant driving scenarios is necessary. Currently, such understanding is obtained by using heuristic algorithms or even by manual inspection of sensor signals. In this paper, we apply deep learning on sensor time series data to automatically extract relevant features for classification of driving scenarios relevant for a Lane-Keep-Assist System. We compare the performance of convolutional and recurrent neural networks and propose two classification models. The first one is an online model for scenario classification during driving. The second one is an offline model for post-processing, providing higher accuracy.

Like all man-made artifacts, Advanced Driver Assistance Systems have certain limitations beyond which they cannot function properly or work at all. The users in the driving domain have diverse backgrounds and it is likely that many of them start interacting with the technology without having gained detailed knowledge, such as by reading user manuals. This study explores whether drivers with only minimal knowledge of the Lane Departure Warning system can recognize the system's speed threshold and update themselves about this in a dynamic driving environment. Since drivers are not always single-mindedly focused on driving so participants were divided into two groups. Group 1 subjects performed only the driving task, while Group 2 was prompted to carry out a secondary task in addition to the primary driving task. Our investigation allowed us to estimate the effects of drivers' minimal mental model of the system working due to lack of sufficient knowledge concerning its capacities upon their 1) learning from experience and 2) situation awareness. It was assumed that the demands of multitasking would impair the driver's ability to observe the system's state and delay the mental model improvement process, compared to a setting in which driving was the only task. Experiment results using a driving simulator have been presented. The results revealed that the number of the drivers, who did not become aware of the operating condition of the LDW system, was quite high for both groups regardless of the nature of their tasks. It had been deduced that the users' preconceptions and expectations due to their limited knowledge about the capacities of the technology, could make them rely on the system to alert them in different situations even when the system was not operating. Hence, these factors can not only adversely affect the social benefits associated with ADAS, but also the acceptance and usability of these systems.