Advanced Driver Assistance Systems Systems Research Articles

Cost reduction for advanced driver assistance systems through hardware downscaling and deep learning

AbstractWith the ever‐growing population density, transportation sector has become haphazard causing accidents and fatalities. It can be reduced by the effective utilization of intelligent transportation systems (ITS). The applications of advanced driver‐assistance systems (ADAS) and deep learning are trending globally with immense research potential and combined with ITS results into an intelligent system with decision‐taking capabilities. High‐priced driving modules are getting introduced day‐by‐day offering humongous processing power and computational capabilities. This research aims at utilizing the downscaled hardware for real‐time vehicular and pedestrian detection using a deep learning algorithm. YOLO v3 deep learning network is incorporated with NVIDIA series of GTX 1060 for real‐time object detection for assisting the ADAS systems. The system offers high precision (0.9618) of object detection in real time with high frame rate (74.36 fps). The comparative analysis between different GPU‐based hardware modules and the proposed module has been carried out keeping in mind the Indian context of automobile usages. The work lays a solid foundation for carrying out research for transportation analysis based on ADAS and deep learning for ITS such as real‐time congestion estimation and accident detection.

Analysis of advanced driver assistance systems in police vehicles: A survey study

Few studies have examined the impact of advanced driver-assistance systems (ADAS) on police officers to improve driver safety and prevent crashes. This is in spite of police officers having higher driving-related mortality rates than average civilians. To fill this gap, a survey study was conducted on 73 police officers to assess their opinions on various ADAS features as well as their recommendations for improvement. Results of the correlation analyses indicated that officer behavior and opinion on ADAS features were influenced by the trust officers had in the available ADAS systems among other key factors such as ADAS training and perceived usefulness. On this basis, guidelines for future research and development of ADAS were provided to improve officer driving safety in police operations. The guidelines need to be further validated in future driving simulation or naturalistic studies.

Machine Learning Based Model Development with Annotated Database for Indian Specific Object Detection

<div class="section abstract"><div class="htmlview paragraph">Now-a-days, Advanced driver-assistance systems (ADAS) is equipping cars and drivers with advance information and technology to make them become aware of the environment and handle potential situations in better way semi-autonomously. High-quality training and test data is essential in the development and validation of ADAS systems which lay the foundation for autonomous driving technology.</div><div class="htmlview paragraph">ADAS uses the technology like radar, vision and combinations of various sensors including LIDAR to automatize dynamic driving tasks like steering, braking, and acceleration of vehicle for controlled and safe driving. And to integrate these advance technologies, the ADAS needs labeled data to train the algorithm to detect the various objects and moments of driver. Image annotation is one the well-known service to create such training data for computer vision.</div><div class="htmlview paragraph">There are number of open source annotated datasets available viz. COCO, KITTI etc. But these datasets are limited to either US or European road environment scenarios. There is hardly any dataset available for Indian Specific road conditions. In order to capture all the India Specific objects on road, we are capturing data through various environmental conditions, road conditions and annotating objects on the collected camera images. 2-D bounding boxes are being used to annotate objects in an image.</div><div class="htmlview paragraph">These annotated images will then be used to train a machine learning model. Finally this trained model will be then used to detect Indian specific objects with real time data acquired through camera sensors mounted on the vehicle. Object detection is one of the prime aspects for ADAS functionality. The complete process will demonstrate end to end solution for object detection techniques from data generation to real-time object identification and recognition.</div></div>

A Sensor Fused Rear Cross Traffic Detection System Using Transfer Learning.

Recent emerging automotive sensors and innovative technologies in Advanced Driver Assistance Systems (ADAS) increase the safety of driving a vehicle on the road. ADAS enhance road safety by providing early warning signals for drivers and controlling a vehicle accordingly to mitigate a collision. A Rear Cross Traffic (RCT) detection system is an important application of ADAS. Rear-end crashes are a frequently occurring type of collision, and approximately 29.7% of all crashes are rear-ended collisions. The RCT detection system detects obstacles at the rear while the car is backing up. In this paper, a robust sensor fused RCT detection system is proposed. By combining the information from two radars and a wide-angle camera, the locations of the target objects are identified using the proposed sensor fused algorithm. Then, the transferred Convolution Neural Network (CNN) model is used to classify the object type. The experiments show that the proposed sensor fused RCT detection system reduced the processing time 15.34 times faster than the camera-only system. The proposed system has achieved 96.42% accuracy. The experimental results demonstrate that the proposed sensor fused system has robust object detection accuracy and fast processing time, which is vital for deploying the ADAS system.

Estimation of Driver Lane Change Intention Based on the LSTM and Dempster–Shafer Evidence Theory

Rapid and correct estimation of driver lane change intention plays an important role in the advanced driver assistance system (ADAS), which could make the driver improve the reliability of the ADAS system and help to decrease driver workload. In this study, a method based on the long short-term memory network (LSTM) and Dempster–Shafer evidence theory is proposed. The model consists of a preliminary decision-making label and a final decision-making label. Driver visual information, head orientation, and vehicle dynamics are collected by preliminary decision-making label. Then, LSTM is used to calculate the initial probability of the driver lane change (left, right, and lane keeping) maneuver intention. The outputs of LSTM are normalized and assigned a basic probability by the Dempster–Shafer evidence theory. The final decision-making label analyzes the information and outputs the probability of each lane change intention and the decision is to identify the driver's current intention. The experimental results show that the accuracy of the model is 90.7% for the intention of changing left and 89.1% for the intention of changing right. The outcome of this work is an essential component for all levels of road vehicle automation.

Overview of Embedded Systems to Build Reliable and Safe ADAS and AD Systems

Automotive industry is a key sector in developed countries, taking advantage from Electronic and Semiconductor industries, for which this work is focused on, including an overview of embedded systems and related technologies for Advanced Driver Assistance Systems (ADAS) development, end user applications and their implementation (SoCs, Application Processors-APs, MCUs, software and boards), manufacturers solutions, architectures, trends and other aspects (like methodologies) to improve functional safety, reliability and performances. The current status to permit the transition from ADAS to Autonomous Driving (AD) systems and Self-Driving Cars (SDC) is also explored.

Real-Time Vehicular Wireless System-Level Simulation

Future automation and control units for advanced driver assistance systems (ADAS) will exchange sensor and kinematic data with nearby vehicles using wireless communication links to improve traffic safety. In this paper we present an accurate real-time system-level simulation for multi-vehicle communication scenarios to support the development and test of connected ADAS systems. The physical and data-link layer are abstracted and provide the frame error rate (FER) to a network simulator. The FER is strongly affected by the non-stationary doubly dispersive fading process of the vehicular radio communication channel. We use a geometry-based stochastic channel model (GSCM) to enable a simplified but still accurate representation of the non-stationary vehicular fading process. The propagation path parameters of the GSCM are used to efficiently compute the time-variant condensed radio channel parameters per stationarity region of each communication link during run-time. Five condensed radio channel parameters mainly determine the FER forming a parameter vector: path loss, root mean square delay spread, Doppler bandwidth, <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">K</i> -factor, and line-of-sight Doppler shift. We measure the FER for a pre-defined set of discrete grid points of the parameter vector using a channel emulator and a given transmitter-receiver modem pair. The FER data is stored in a table and looked up during run-time of the real-time system-level simulation. We validate our methodology using empirical measurement data from a street crossing scenarios demonstrating a close match in terms of FER between simulation and measurement.

Pre-crash control strategy of driver assistance system

The aim of the article is the optimisation process of the ADAS (Advanced Driver Assistance Systems) control. The methodology is based on the classi-fication of ADAS systems according to the situations of unavoidable accidents. Theevaluation of expected consequences uses injury biomechanics, which represents theextended definition of HMI (Human-Machine Interaction). The evaluation of injury mechanism and the machine intervention enables to control this process withthe target to minimise the consequent injuries. Then the decision making takesnew inputs to the control process and it enriches the multiparametric control ofthe system with the target to minimise the consequences.

СТРУКТУРА И ИЕРАРХИИ ИНФОКОММУНИКАЦИОННЫХ СИСТЕМ В СИСТЕМАХ ПОМОЩИ ВОДИТЕЛЮ КАК ШАБЛОНЫ ПРОЕКТИРОВАНИЯ ПОДСИСТЕМ БЕСПИЛОТНЫХ ТРАНСПОРТНЫХ СРЕДСТВ

The article presents the structure of infocommunication systems of unmanned transport systems developed by the author, which combines the main information subsystems of the control loop in Advanced driver-assistance systems (ADAS) and shows their interaction. The analysis of the architecture of the ADAS system as a complex system from the point of view of the application of Mesarovich's theory of hierarchical multilevel systems is given (also known as stratification process) - as a result of which ADAS infocommunication systems are presented in the form of a hierarchy of the strata-layers-echelons type. Such a representation has scientific novelty as the application of stratification methods to the description of ADAS infocommunication systems and practical value in the design of the ADAS architecture and the creation of templates for the design of unmanned transport systems infocommunication subsystems.

Deriving functional safety (ISO 26262) S-parameters for vulnerable road users from national crash data

Currently, advanced driver assistance systems (ADAS) and automated vehicles (AV) are designed for use in the existing road infrastructure. These partially and fully automated vehicles will be operated in a shared space not only with other vehicles but also with vulnerable road users (VRU). Even though crashes between ADAS equipped vehicles or AV and VRU seem inevitable in such a scenario, functional safety, i.e., the assessment of the quality and safety level of the automation system, plays a crucial role in minimizing the crash frequency and the injury severity. We develop a data-driven approach to injury severity estimation for functional safety, i.e., ISO 26262 S-parameters, for four types of VRU: pedestrians, bicyclists, scooterists, and motorcycle riders. To estimate the S-parameter, the 90th-percentile of the injury severity distribution in the S-scale, a population-based data set (Germany's national data set DESTATIS) is used. Since the description of the injury severity in DESTATIS is not detailed enough for a direct one-to-one mapping to the S-scale, we enhance the level of detail in the population-based data set by using additional information from the German in-depth accident study (GIDAS), an in-depth, size-limited survey of part of the same population. Thus, we are able to transform the 4-level injury scale (uninjured, slightly injures, severely injured, and fatal) of the police reports found in DESTATIS into the three breakpoints of the injury severity scale (ISS) (ISS ≥{4, 9, 16}) which in turn directly translate to the four levels of the S-scale. Furthermore, the ISS ≥9 breakpoint more or less equates to MAIS 3+, the definition of ‘severe injury’ in nearly all international road safety goals that look beyond fatalities. The derived injury scale transformation can be utilized to translate the injury severities of the police-reported cases to the politically needed MAIS 3+ distribution. Thus, population-based data can be directly used to estimate the proportion of these ‘severely injured.’ The crashes are analyzed from the perspective of the VRU as well as from the vehicle type involved. We stratified the opposing vehicles by injury mechanism: wrap projection for bonnet type passenger vehicles (BTV), forward projection for box type vehicles like light trucks (LTV), as well as single-vehicle crashes. We cluster the crash data into traffic domains based on the speed limit: shared zone, residential streets, city roads, arterial thoroughfares, rural roads, and autobahn. For each VRU type, injury mechanism, and traffic domain, the S-parameters, i.e., the 90th-percentile of the injury severity measured in S-scale, are calculated with a one-sided 95% confidence level. Exemplary applications of the results are given in the discussion: an evaluation of an AV hitting a crossing pedestrian, an in-lane swerving ADAS system for VRU avoidance, and the rating of the nominal performance of an inflatable helmet for pedestrians.

Acceptance of Transportation Technologies by Aging Adults

Older adults are at greater risk of death and serious injury in transportation crashes which have been increasing in older adult cohorts relative to younger cohorts. Can technology provide a safer road environment? Even if technology can mitigate crash risk, is it acceptable to older road users? We outline the results from several studies that tested 1) whether advanced driver assistance systems (ADAS) can improve older adult driving performance, 2) older adults’ acceptance of ADAS and Autonomous Vehicle (AV) systems, and 3) perceptions of value for ADAS systems, particularly for blind-spot detection systems. We found that collision avoidance warning systems improved older adult simulator driving performance, but not lane departure warning systems. In a young to middle-aged sample the factor “concern with AV” showed age effects with older drivers less favorable. Older drivers, however, valued an active blind spot detection system more than younger drivers.

Acceptance and Uptake of Technology Among Older Adults: Findings From CREATE

Technology is ubiquitous in everyday life and technology use is essential to independent living. Technology applications are deployed in most settings (e.g., healthcare, work) and also changing how we communicate, and access information and services. Despite increases in technology uptake among older people, an age-related digital divide remains. Drawing from research conducted by the Center for Research and Education on Aging and Technology Enhancement (CREATE), this symposium will discuss the acceptance and uptake of technology among older adults and factors influencing technology adoption. S. Czaja will discuss recent technology trends and how they vary according to technology type and subgroups of older adults. Based on CREATE findings, she will also discuss age group and cohort differences in interest in and comfort with technology. W. Boot will present findings from a study that examined adherence to a technology-based cognitive training program and how individual difference factors shaped adherence. N. Charness will present findings regarding whether advanced driver assistance systems (ADAS) can improve older adult driving performance, and older adults’ acceptance and perceptions of value of ADAS systems. M. Harris will discuss health related technology interventions and how integration of technology acceptance and behavior change models can provide insights for the design of health behavior interventions aimed at older adults. J. Sharit will provide findings from a study, which examined the willingness of older adults to adopt a variety of technologies and factors influencing willingness to adopt. Bo Xie will lead a discussion of these issues and outline areas for future research.

Study of the influence of the exhaust line ultrasounds over the performance of the Blind Spot Warning System

Abstract During the previous years, the vehicle manufacturers have tried to equip their vehicles with as much technology as possible, making the driving experience for people easier than ever. Most of the modern vehicles come today with ADAS (Advanced Driver Assistance Systems) either for driving (E.g. Cruise Control, Blind Spot Warning) or Parking (E.g. Rear Ultrasonic Sensors, Rear View Camera). Since the vehicle come equipped with more technology, a major task in developing vehicle remains the integration of these ADAS system in the vehicle context with the other components. Since most of the components cope with each other on the vehicle level, some technologies are more affected by other components – such as the case of an ultrasound vehicle scanning system (Blind Spot Warning) and the Exhaust line that emits ultrasounds from the exhaust muffler. The aim of this paper is to study the influence of the exhaust line ultrasounds (ultrasounds that are emitted by the engine cycle and filtered in the exhaust line of the vehicle) over the detection performance of the Blind Spot Warning Ultrasound system. Since vehicles are sold with a wide variety of powertrains, the solution presented took into account also these differences between powertrains equipped. In order to test the solution, mock-ups of the vehicle were made in order to proof the robustness of the method.

Vehicle Detection Based on Multifeature Extraction and Recognition Adopting RBF Neural Network on ADAS System

A region of interest (ROI) that may contain vehicles is extracted based on the composite features on vehicle’s bottom shadow and taillights by setting a gray threshold on vehicle shadow region and a series of constraints on taillights. In order to identify the existence of target vehicle in front of Advanced Driver Assistance System (ADAS) for the extracted ROI, a neural network recognizer of the Radial Basis Function (RBF) is found by extracting a series of parameters on the vehicle’s edge and region features. Using a large amount of collected images, the ROI that may contain vehicles is verified to be effective by extracting composite features of the shadow at the bottom of vehicle and taillights. Based on the positive and negative sample base of vehicles, the neural network recognizer is trained and learned, which can quickly realize network convergence. Furthermore, the vehicle can be effectively identified in the region of interest using the trained network. Test results show that the vehicle detection method based on multifeature extraction and recognition method based on RBF network have stable performance and high recognition accuracy.

Gradient-Based Edge Effects on Lane Marking Detection using a Deep Learning-Based Approach

Lane detection is part of the advanced driver assistance system (ADAS) equipped in intelligent vehicles. The system provides the driver with significant geometric information of the road ahead. Numerous deep learning techniques have been employed in lane detection because of the simplicity, ease, and efficiency of these techniques in learning discriminative features from RGB (red, green, and blue) images. However, existing works have rarely considered detecting lane markings during bad weather conditions, which could reduce lane detection performance. Hence, this paper proposed a Fully Convolutional Network (FCN) model with RGB and Canny edge detection used as the model’s spatial input. The proposed platform was developed using two scenarios: FCN-RGB-edge and FCN-edge. The model development was divided into three stages, namely data acquisition, platform development, and benchmarking against existing methods and data. Both scenarios using the proposed method yielded a 4% improvement compared to the original FCN-RGB images (i.e., the previous method). The Canny edge detection method successfully extracted necessary information from the images and neglected the water drops in rainy conditions by treating them as noise. In summary, the proposed method has the potential to boost the performance of the ADAS system in detecting lane markings in rainy conditions.

A Study on the Evaluation Method of Autonomous Emergency Vehicle Braking for Pedestrians Test Using Monocular Cameras

Traffic accidents continue to increase in Korea as traffic increases, and the resulting loss of life is also on the rise. According to data surveyed by the South Korean National Police Agency, 45,921 pedestrian traffic accidents were reported in 2019, resulting in 1487 deaths and 46,400 injuries. Due to the increased interest in traffic accident safety, the Advanced Driver Assistance System (ADAS) concept is rapidly developing and playing a significant role in coping with activities that are not recognized by the driver. Autonomous Emergency Braking (AEB), a representative ADAS system, is a system that is useful for preventing and mitigating accidents by braking vehicles in emergencies. For the study of AEBs’ safety evaluation methods for pedestrians, a distance measurement method using a monocular camera with excellent accessibility, and measurement equipment required to validate data on the movement of vehicles, and a dummy to replace pedestrians are used. Based on the evaluation scenario considering the proposed Korea road environment, the relative distance obtained from equipment like DGPS and the relative distance using a monocular camera is compared and analyzed to verify safety. Comparative analysis shows that the minimum deviation is 2.3 cm, the third test result of 30 km/h of Car-to-Pedestrian Nearside Child (CPNC), and the maximum deviation is 25 cm, the first test result of 25 km/h of Car-to-Pedestrian Nearside Adult (CPNA). The main factor in error generation is that the lane recognition in the camera image is not accurate, and the perception of small children is slow, which is why emergency braking is considered to have been slow. It is deemed that a safety assessment in weather conditions of adverse conditions will be required in the future.

Anomaly Detection Based on RBM-LSTM Neural Network for CPS in Advanced Driver Assistance System

Advanced Driver Assistance System (ADAS) is a typical Cyber Physical System (CPS) application for human–computer interaction. In the process of vehicle driving, we use the information from CPS on ADAS to not only help us understand the driving condition of the car but also help us change the driving strategies to drive in a better and safer way. After getting the information, the driver can evaluate the feedback information of the vehicle, so as to enhance the ability to assist in driving of the ADAS system. This completes a complete human–computer interaction process. However, the data obtained during the interaction usually form a large dimension, and irrelevant features sometimes hide the occurrence of anomalies, which poses a significant challenge to us to better understand the driving states of the car. To solve this problem, we propose an anomaly detection framework based on RBM-LSTM. In this hybrid framework, RBM is trained to extract general underlying features from data collected by CPS, and LSTM is trained from the features learned by RBM. This framework can effectively improve the prediction speed and present a good prediction accuracy to show vehicle driving condition. Besides, drivers are allowed to evaluate the prediction results, so as to improve the accuracy of prediction. Through the experimental results, we can find that the proposed framework not only simplifies the training of the entire neural network and increases the training speed but also greatly improves the accuracy of the interaction-driven data analysis. It is a valid method to analyze the data generated during the human interaction.

Determination of Fast-Moving Objects’ Speed and Range with Linear Frequency Modulation Continuous Wave Radar Using Autocorrelation Scheme

Introduction. A hardware basis of modern Advanced Driver Assistance Systems (ADAS) consists of millimeterrange radars, characterized by a relatively short range (meters – tens of meters). At the same time, improving of traffic safety requires to increase the range at least to several hundred meters. The one way to achieve such values is to increase wavelength of a probing signal, to use the centimeter range of wavelengths, for example. The paper represents a detailed description of main steps of signal processing algorithm in the model of the ADAS low-power centimeter range radar, which provides fast-moving objects speed and range definition.Aim. Development of an algorithm for estimating the range and the speed of targets by an autocorrelation radar with a wide-band continuous linear frequency modulation (linear FM) signal in order to increase the rate of the ADAS system estimates formation.Materials and methods. The proposed algorithm is based on the methods of primary and secondary digital processing of radar signals. The model of a centimeter-range autocorrelation radar with a broadband continuous linear FM probing signal was used for practical researches. MATLAB software was used to process the received signal samples.Results. The algorithm has been developed to determine the speed and the range of fast-moving objects in conditions when their movement during the evaluation interval significantly exceeds the radar range resolution. The use of simplified Kalman filtering for inter-period secondary signal processing allowed to increase significantly the stability of the algorithm. In a full-scale experiment using the low-power radar model with continuous radiation of the centimeter range, it was shown that a stable assessment of a real car speed and range was provided at a distance of at least about one kilometer.Conclusion. The results of the field experiment make it possible to draw conclusions that the proposed algorithm is highly robust even in the absence of inter-period secondary processing. Its usage allows one to improve the stability of the algorithm without considerable additional computational costs. It is possible because near-linear dynamics of the observation object and of the radar carrier makes it sufficient to use a simplified implementation of Kalman filter in the form α-β-algorithm.

Resource-Constrained Machine Learning for ADAS: A Systematic Review

The advent of machine learning (ML) methods for the industry has opened new possibilities in the automotive domain, especially for Advanced Driver Assistance Systems (ADAS). These methods mainly focus on specific problems ranging from traffic sign and light recognition to pedestrian detection. In most cases, the computational resources and power budget found in ADAS systems are constrained while most machine learning methods are computationally intensive. The usual solution consists in adapting the ML models to comply with the memory and real-time (RT) requirements for inference. Some models are easily adapted to resource-constrained hardware, such as Support Vector Machines, while others, like Neural Networks, need more complex processes to fit into the desired hardware. The ADAS hardware (HW platforms) are diverse, from complex MPSoC CPUs down to classical MCUs, DPSs and application-specific FPGAs and ASICs or specific GPU platforms (such as the NVIDIA families Tegra or Jetson). Therefore, there is a tradeoff between the complexity of the ML model implemented and the selected platform that impacts the performance metrics: function results, energy consumption and speed (latency and throughput). In this paper, a survey in the form of systematic review is conducted to analyze the scope of the published research works that embed ML models into resource-constrained implementations for ADAS applications and what are the achievements regarding the ML performance, energy and speed trade-off.

Pedestrian Detection in Severe Weather Conditions

Pedestrian detection has never been an easy task for computer vision and the automotive industry. Systems like the advanced driver-assistance system (ADAS) highly rely on far-infrared (FIR) data captured to detect pedestrians at nighttime. The recent development of deep learning-based detectors has proven the excellent results of pedestrian detection in perfect weather conditions. However, it is still unknown what the performance in adverse weather conditions is. In this paper, we introduce a 16-bit thermal data dataset called ZUT (Zachodniopomorski Uniwersytet Technologiczny) as having the widest variety of fine-grained annotated images captured in the four biggest European Union countries captured during severe weather conditions. We also provide a synchronized Controller Area Network (CAN bus) data, including driving speed, brake pedal status, and outside temperature for future ADAS system development. Furthermore, we have tested and provided 16-bit depth modifications for the YOLOv3 deep neural network (DNN) based detector, reaching a mean Average Precision (mAP) up to 89.1%. The ZUT dataset is published and publicly available at IEEE Dataport and Github.