Browse Topic: Lane keeping assistance
The increase of autonomy demand in the automotive industry made the usage of AI models inevitable. However, such models introduce a variety of threats to automobile safety and security. ISO/PAS 21448 SOTIF is a safety standard that is designed to deal with risks due to non-electrical and non-electronic failures. In this paper we put SOTIF into practice. In our work we introduce a conceivable safety critical scenario that targets the lane keep assist function. We use the suggested modelling techniques in the SOTIF standard to analyze the scenario and extract the trigger event. In result, we propose a contextual based predictive ML model to monitor the intervention between the driver and lane keep assist system. Our approach followed the SOTIF verification and validation guidelines. Empirically, we use a real safety critical scenario dataset as well as an augmented dataset. Our results show a high precision/recall values that exceed 90% by an increase of more than 150% in f1 score compared to non contextual models. It also showed that there is a trade-off relation between the precision/recall values and the sensitivity of models to its inputs.
Implementing Advanced Driver Assistance Systems (ADAS) features that are available in all road scenarios and weather conditions is a big challenge for automotive companies and considered key enablers to achieve autonomous Level 4 (L4) vehicles. One important feature is the Lane Keep Assist System (LKAS). Most LKAS systems are based on lane line detection cameras and lane coefficient estimations by the camera is the key point for LKAS where the camera recognizes the lane lines using edge detection. But when the lane markers are not available due to high traffic and slow driving on the roads, another source of data for the lane lines needs to be available for the LKAS. In this paper a multi-sensor fusion approach based on camera, Lidar, and GPS is used to allow the vehicle to maintain its lateral location within the lane. The lateral distances of the lane lines are measured by LiDAR detection of the markers based on the intensity and fused with lane line information from the HD Map after transforming the sensors to the same reference. This approach was tested on FEV’s Smart Vehicle Demonstrator and the test results show the vehicle was able to maintain the lane.
A veteran tester who has datalogged many thousands of miles in the U.S. and Japan offers suggestions for rapidly acquiring good test and validation data. Testing for advanced driver-assist systems (ADAS) has required a completely new approach to testing. The most obvious reason for this is the sheer number of sensors and actuators involved in any given feature. Whereas engineers used to test single sensors at a time, today there are upward of 20 sensors to specify. These sensors - short- and long-range radar, mono/stereo cameras, sonar and lidar - need to act in concert. This means that several layers of sensor fusion must be implemented. For example, when it comes to the actuators for the braking system, there are up to seven distinct systems able to apply the brakes. As a result, a failure is not easily tracked to a single root cause but can be a complex combination of several factors. Imagine expanding this to ADAS and autonomous-vehicle (AV) functionalities such as automated cruise control (ACC), lane-keeping assist (LKA), automated emergency braking (AEB) and valet parking, as well as convenience features like blind-spot monitoring, night vision and steering beams, to name just a few.
Lane keeping assist system (LKAS) is an advanced assistant driving system, which can effectively avoid unconscious lane departures caused by drivers due to distraction, fatigue driving, insensitive response to emergencies, etc., reduce traffic accidents, and effectively improve Driving safety. This paper is committed to the research of LKAS standard of commercial vehicles in China. According to the working principle of LKAS and the test and evaluation methods of LKAS at home and abroad, the LKAS test scheme of commercial vehicles was designed after repeated discussions and demonstrations in many meetings, including straight and curve test scenarios. The actual vehicle test of commercial vehicle LKAS were completed by CATARC Automotive Test Center (Tianjin) Co., Ltd. in CATARC Yancheng Automotive Proving Ground Co., Ltd. The performance of LKAS of freight cars and passenger cars under different loads was studied. Based on the test results, suggestions on the formulation of domestic LKAS standards are put forward. This paper not only improves the domestic LKAS test method and evaluation system, but also provides reference for the formulation of domestic LKAS national standards.
In recent decades, research and development in the field of autonomous vehicles have rapidly increased throughout the world, and autonomous driving technologies have begun to be applied to mass production vehicles. Especially recently, even affordable mass production vehicles have begun to be equipped with some autonomous driving systems such as a Lane Keeping Assist (LKA) system. In general, mass-produced LKA systems use a lane detection camera as a means of keeping the lane. One of the common limitations of camera-based LKA systems is that the lane keeping performance significantly decreases when the camera cannot detect lane markings for various reasons such as snow coverage or blurred lane markings. To overcome this limitation, we have developed Global Navigation Satellite System (GNSS)-based LKA systems, which are not affected by the surrounding environment such as weather and the condition of lane markings. In our latest study, we applied Model Predictive Control (MPC) to our GNSS-based LKA system so as to enhance lane-keeping performance. We then revealed that the GNSS-based LKA system with MPC had low robustness regarding the time delay of a GNSS and that countermeasures for the time delay were necessary. In this paper, we apply Smith predictor-like Time Delay Compensation (TDC) to compensate for the time delay. The TDC predicts the current state variables from the past sensor signals based on the vehicle dynamics. We demonstrate that the TDC stabilizes the LKA system even when the GNSS has a time delay in a simulation. Furthermore we add another TDC to compensate for the time delay of Electrical Power Steering (EPS) with the aim of reducing the oscillation of the steering wheel angle. Finally, we evaluate the lane keeping performance in a real-vehicle experiment on a snow-covered highway.
Advanced Driver Assistance Systems (ADAS) like Lane Departure Warning (LDW) and Lane Keep Assist (LKA) have been available for several years now but has experienced low customer acceptance and market penetration. These deficiencies can be traced to the inability of many of the perception systems to consistently recognize lane markings and localize the vehicle with respect to the lane markings in the real-world with poor markings, changing weather conditions and occlusions. Currently, there is no available standard or benchmark to evaluate the quality of either the lane markings or the perception algorithms. This work seeks to establish a reference test system that could be used by transportation agencies to evaluate the quality of their markings to support ADAS functions that rely on pavement markings. The test system can also be used by designers as a benchmark for their proprietary systems. To support this development, an extensive video dataset was collected at different times of day and weather conditions on various roads in Central Texas. The videos were evaluated on different state-of-the art lane detection algorithms and their performance was ranked based on a set of metrics specifically developed for evaluating the effectiveness of the lane estimation system. The test scenarios are comprised of a set of roadways and environmental features, as well as the pavement marking presence and luminance variables. A systems approach is presented by correlating the algorithm performance data to the environmental factors, lane marking types, color, material, and the retroreflectivity of pavement markings.
Lane keeping assist (LKA) is an autonomous driving technique that enables vehicles to travel along a desired line of lanes by adjusting the front steering angle. Reinforcement learning (RL) is one kind of machine learning. Agents or machines are not told how to act but instead learn from interaction with the environment. It also frees us from coding complex policies manually. But it has not yet been successfully applied to autonomous driving. Two control strategies using different deep reinforcement learning (DRL) algorithms have been proposed and used in the lane keeping assist scenario in this paper. Deep Q-network (DQN) algorithm with discrete action space and deep deterministic policy gradient (DDPG) algorithm with continuous action space have been implemented, respectively. Based on MATLAB/Simulink, deep neural networks representing the control policy are designed. The environment as well as the vehicle dynamics are also modelled in Simulink. By integrating the proposed control method and a vehicle dynamics model, the lane keeping assist simulation is performed. Experimental results demonstrate that the vehicle travel along the centerline of the path and the controller reaches a steady state after a short time, validating the effectiveness of the proposed control method.
Together with its many partners, ZF supplies camera and radar technology and advanced components for both the passenger car and commercial truck markets, the latter being especially suited for the move to more complex driver-assistance systems, according to Dan Williams, director of ADAS & Autonomy at ZF. “The business case in commercial vehicle for reduction in driver hours of service, fuel cost reduction and safety have strong economic incentives to adopt ADAS/automated driving technology. Additionally, the regulations placed on the industry will require our customers to utilize certain solutions,” he said. ZF is working on both highly automated “revolutionary” systems and on “evolutionary” driver-assistance systems that are increasingly complex, he said, citing the supplier's OnTraX lane keep assist that will launch in 2020 with its first major OEM customer. Williams spoke with TOHE at the recent NACV Show in Atlanta, and he's scheduled to participate in a Commercial Vehicle Safety technical session at the SAE Government/Industry Meeting taking place January 22-24, 2020, in Washington, DC (www.sae.org/attend/government-industry/).
Lane Keeping Assistance System (LKAS) is a typical lateral driver assistance system with low acceptance. One of the main reasons is that fixed parameters cannot satisfy individual differences. So LKAS adaptive to driver characteristics needs to be designed. Driver Steering Override (DSO) process is an important process of LKAS. It happens when contradiction between driver’s intention and system behavior occurs. As feeling of overriding will affect the overall experience of using LKAS, the design of DSO characteristics is worthy of attention. This research provided an adaptive design scheme aiming at DSO characteristics for LKAS by building Driver Preference Model (DPM) based on simulator test data from preliminary experiments. The DPM was to represent the relationship between driver characteristics indices and driver preferred system characteristics indices. So that new drivers’ preference can be predicted by DPM based on their own daily driving data with LKAS switched off. The inputs of DPMs are 27 lane changing driver characteristics indices which were extracted based on natural lane changing data. Principal Component Analysis (PCA) and correlation analysis were used during input-indices selection. The outputs of DPMs are 2 driver preferred system characteristics indices which can express driver’s preferred DSO characteristics. 6 participants took part in preliminary experiments for data collection and evaluation experiments. Their preferred system characteristic indices were calculated through the results of evaluation experiments. The building of DPMs was realized by Random Forest (RF) based on 6 participants’ lane changing driver characteristics indices and driver preferred system characteristics indices. The verification of DPMs’ effectiveness was realized by subjective evaluation. The result showed that DPMs can realize adaptive design of DSO characteristics for LKAS focusing on driver’s preference and improve driver’s acceptance.
Objectives: The project goal was to create an initial set of standardized tests to explore whether they enable the ongoing evaluation of automated driving features as they evolve over time. These tests focused on situations that were representative of several daily driving scenarios as encountered by lower-level automated features, often called Advanced Driver Assistance Systems (ADAS), while looking forward to higher levels of automation as new systems are deployed. Methods: The research project initially gathered information through a review of existing literature about ADAS and current test procedures. Thereafter, a focus group of industry experts was convened for additional insights and feedback. With this background, the research team developed a series of tests designed to evaluate a variety of automated driving features in currently available implementations and anticipated future variants. Key ADAS available on current production vehicles include adaptive cruise control (ACC), lane keeping assist (LKA), and automatic emergency braking (AEB). Seven of the most automated production vehicles available in 2018 from six manufacturers were subjected to a series of standardized tests that were performed on a closed test track environment to assess the vehicle capabilities and limitations of the automated driving systems’ operational domains. Results: Considerable performance variability was observed between different vehicle manufacturers and within a single vehicle model across repeated trials and multiple replications. In addition, there were specific roadway characteristics that significantly impacted performance. Conclusions: The results indicate that standardized testing can assist researchers in determining the current capabilities of vehicles with automated driving features. The research team suggests continuing to improve and expand standardized testing of automated driving features and to work toward industry consensus of a robust evaluation mechanism that may play a key role in the conformance of future automated-vehicle systems.
Lane Keeping Assistance (LKA) system is a very important part in Advanced Driver Assistance Systems (ADAS). It prevents a vehicle from departing out of the lane by exerting intervention. But an inappropriate performance during LKA intervention makes driver feel uncomfortable. The intervention of LKA can be divided into 3 parts: intervention timing, intervention process and intervention ending. Many researches have studied about the intervention timing and ending, but factors during intervention process also affect driver feelings a lot, such as yaw rate and steering wheel velocity. To increase driver’s acceptance of LKA, objective and subjective tests were designed and conducted to explore important indices which are highly correlated with the driver feelings. Different kinds of LKA controller control intervention process in different ways. Therefore, it’s very important to describe the intervention process uniformly and objectively. This paper proposes 16 Characteristic Indices (CI), such as ‘maximum yaw rate’, to describe steering wheel motion, vehicle motion and other aspects during variable LKA intervention processes. Then, to acquire drivers’ subjective evaluation about LKA, a questionnaire including 3 questions from different aspects was designed for drivers to give Subjective Ratings (SR). Lastly, to describe the nonlinear correlation between CI and SR, Random Forests (RF) algorithm was used to establish the correlation model. Different from other modeling methods, RF can not only build the model by data training, but also give out the importance of each CI in the model. Through this method, important indices really affecting the driver feelings during the LKA intervention process were explored. What’s more, by the use of CI, the correlation model can predict the driver feelings regardless of specific LKA controller type, thus important indices can be optimized, which means the prediction about SR can be used to offer necessary guidance to the development and calibration of LKA system.
Lane-keeping assist system (LKA) alerts the driver or intervenes in the driving when the vehicle deviates from the lane. But its effect is highly dependent on the driver’s acceptance. Distance to Lane Crossing (DTLC) and Time to Lane Crossing (TTLC) are two important factors to consider the danger level of the scenario, which are also two references for drivers to make decisions. At present, most of the functional design standards are based on these values, while they often differ for different vehicle movements. This study uses a driving robot to precisely control the test conditions and performs field tests on two advanced autonomous vehicles in National Intelligent Connected Vehicle (Shanghai) Pilot Zone. The test conditions are extended based on various test standards and the LKA performance of vehicles in the pre-experiment. The application of high-precision maps and RT systems in the test provided positioning information for the driving robot with an accuracy error of less than 2 cm. The FIR filtering of the audio and video data is used to obtain the driver’s feedback on the alarm or intervention. According to the UTC time synchronization, the thresholds of the DTLC and TTLC at that moment are obtained. Finally, one-way ANOVA method is used to obtain the DTLC or TTLC distribution of the vehicle movement status characteristics. This result can be used to compare with the analysis of natural driving behavior, and then apply in proposing a more reasonable Human Machine Interaction (HMI) scheme considering the driver’s operating characteristics.
In order to satisfy design requirements of Lane Keeping Assistance System (LKAS), a Driver Steering Override (DSO) strategy is necessary for driver’s interaction with the assistance system. The assistance system can be overridden by the strategy in case of lane change, obstacle avoidance and other emergency situations. However, evaluation and optimization of the DSO strategy for LKAS cannot easily be completed quantitatively considering driver’s acceptability. In this research, firstly subjective and objective evaluation experiment is designed. Secondly, correlations between the subjective and the objective evaluation results are established by using regression analysis. Finally, based on the correlations established previously, the optimal performance of DSO strategy is obtained by setting the desired comprehensive evaluation ratings as the optimized goal. Except for the whole process of the research, there are some details of the subjective and objective evaluation experiment design to be introduced. For the objective evaluation experiment, several objective characteristic indices (CI) are extracted, which are unrelated to controller. As for the subjective evaluation experiment, a questionnaire consisting of 7 questions is initially designed. Then, based on pre-experiment results, it is simplified by correlation analysis among its questions to eliminate information redundancy. Additionally, proper weights are given to different questions of the questionnaire by using Analytic Hierarchy Process (AHP), and the driver’s comprehensive evaluation of DSO strategy is added on by this method. Overall, with the conduct of the newly designed experiment, it would be possible to achieve the following five advantages. The first one is a generic objective evaluation regardless of controller types. The second one is a more simplified subjective evaluation experiment. The third one is a reliable comprehensive subjective evaluation. The forth one is the optimization of DSO strategy for LKAS based on the driver’s acceptability. The last one is the obtained target objective CI can be used for different types of controller design.
Recently, development of vehicle control system targeting Full Driving Automation (autonomous driving level 5) has advanced. Some applications of autonomous driving systems like the Lane Keeping Assist system (LKA) and Auto Lane Change system (ALC) (autonomous driving level 1-3) have been put on the market. However, the conventional system using information from front camera, it is difficult to operate in some situations. For example the road that no line, large curvature and number of lane increases or decreases. We propose an autonomous driving system using high accuracy vehicle position estimation technology and a high definition map. An LKA system calculates the target steering wheel angle based on both vehicle position information from the Global Navigation Satellite System (GNSS) and the target lane of high the definition map, according to the method of front gaze driver model. Then, the system controls steering the wheel angle by Electric Power Steering (EPS). In the case of ALC, a target lane-change path is generated based on information of the vehicle’s own lane and the next one. The proposed system solved the problem of the conventional method. Moreover, the developed method can operate more smoothly than the conventional one. Finally, we demonstrate that the proposed system enables the actual vehicle to operate LKA and ALC in the merged road of a test tracks.
Emerging autonomous driving technologies, with emergency navigating capabilities, necessitates innovative vehicle steering methods for operators during unanticipated scenarios. A reconfigurable “plug and play” steering system paradigm enables lateral control from any seating position in the vehicle’s interior. When required, drivers may access a stowed steering input device, establish communications with the vehicle steering subsystem, and provide direct wheel commands. Accordingly, the provision of haptic steering cues and lane keeping assistance to navigate roadways will be helpful. In this study, various steering devices have been investigated which offer reconfigurability and haptic feedback to create a flexible driving environment. A joystick and a robotic arm that offer multiple degrees of freedom were compared to a conventional steering wheel. To evaluate the concept, human test subjects interacted with the experimental system featuring a driving simulator with target hardware, and completed post-test questionnaires. Based on the data collected, drivers’ lane keeping performance was superior using a haptic robotic arm with haptic feedback to the joystick and steering wheel with an improvement of up to 70.18% during extreme maneuvers. Haptic feedback, with a lane keeping algorithm, can assist the operator in steering the vehicle given the likely deterioration of driving skills when autonomous vehicles become prevalent.
RACam [1] is an Active Safety product designed and manufactured at Delphi and is part of their ADAS portfolio. It combines two sensors - Electronically Scanned RADAR and Camera in a single package. RADAR and Vision fusion data is used to realize safety critical systems such as Adaptive Cruise Control (ACC), Autonomous Emergency Braking (AEB), Lane Departure Warning (LDW), Lane Keep Assist (LKA), Traffic Sign Recognition (TSR) and Automatic Headlight Control (AHL). Figure 1 RACam Front View. With an increase in Active Safety features in the automotive market there is also a corresponding increase in the complexity of the hardware which supports these safety features. Delphi’s hardware design for Active Safety has evolved over the years. In Delphi’s RACam product there are a number of critical components required in order to realize RADAR and Vision in a single package. RACam is also equipped with a fan and heater to improve the operating temperature range. RADAR and Camera sensors go through unit specific calibration to adjust the sensitivity and alignment of the sensors. Manufacturability of this advanced system needs to be part of the design solution. To achieve flawless manufacturing of RACam, Delphi uses XCP protocol at different stages of the manufacturing process. XCP commands are used to verify RADAR, Camera, Heater, Fan, Memory Integrity and Diagnostics. RADAR and Camera sensors are calibrated using XCP commands and calibrations are stored in flash memory. These calibrations are also used for sensor alignment in the vehicle. This paper provides design details of how the XCP protocol is used in the RACam project. RACam specific functions are realized using XCP user commands and they will be discussed in this paper.
Mitsubishi Electric has been developing a lane keeping assist system (LKAS). This system consists of our products such as an electric power steering (EPS), a camera, and an electronic control unit (ECU) for ADAS. In this system, the camera detects a lane marker, the ECU estimates reference path and vehicle position, and calculates reference steering wheel angle, and the EPS controls a steering wheel angle based on reference steering wheel angle. In this paper, we explain the calculation method of reference steering wheel angle for path tracking control. We derive a formula of reference steering wheel angle calculation that converges lateral position deviation in desired time by using lateral position deviation change rate control on forward gaze point as path tracking control algorithm. Since the formula is obtained from the vehicle model, we can easily design a controller depending on the vehicle type, by using known vehicle specifications. In addition, we confirmed that the algorithm can be used for LKAS by simulation and examination. The algorithms uses just the lateral position deviation, yaw angle deviation on the forward gaze point and the yaw rate. Since, the algorithm doesn’t need the road curvature information, it can be used for applications such as path tracking control with global navigation satellite system.
ABSTRACT In any active safety system, it is desired to measure the “performance”. For the estimation case, generally a cost function like Mean-Square Error is used. For detection cases, the combination of Probability of Detection and Probability of False Alarm is used. Scenarios that would really expose performance measurement involve complex, dangerous and costly driving situations and are hard to recreate while having a low probability of actually being acquired . Using a virtual tool, we can produce the trials necessary to adequately determine the performance of active safety algorithms and systems. In this paper, we will outline the problem of measuring the performance of active safety algorithms or systems. We will then discuss the approach of using complex scenario design and Monte Carlo techniques to determine performance. We then follow with a brief discussion of Prescan and how it can help in this endeavor. Finally, two Monte Carlo type examples for particular active safety algorithms (LDW and AEB) will be presented.
A novel speed and position dependent Lane Keeping Assistance (LKA) control strategy for heavy vehicles is proposed. This LKA system can be implemented with any torque overlay system capable of accepting external position or torque commands. The proposed algorithm tackles the problem of lane keeping in two ways from a heavy vehicle's perspective. First, it stabilizes the vehicle's lateral position by bringing it to the center of the lane and giving it the correct heading to stay there. This is done using a speed and position dependent control strategy that becomes less aggressive as the vehicle's speed increases and as it gets closer to the center of the lane. Such speed and position dependency is especially critical in heavy vehicles where unnecessary aggressive control can lead to oscillations about the lane's centerline when cruising at high speeds. Furthermore, the proposed controller allows the vehicle to negotiate the road's curvature efficiently while tracking the lane's centerline. This is achieved using a feed-forward strategy based on the angle of attack needed to negotiate a road of a particular curvature at a particular speed. Ultimately, the new LKA system was implemented into a torque overlay system [1, 2], and tested on a heavy vehicle. As a result, significant improvement in lane center tracking was noted, as well as in negotiating road curvature. These capabilities are expected to make driving heavy vehicles such as tractor-trailers and motor homes less strenuous, and have the potential to be the basis for autonomous heavy vehicle applications.
Predicting driver response to road departure and attempted recovery is a challenging but essential need for estimating the benefits of active safety systems. One promising approach has been to mathematically model the driver steering and braking inputs during departure and recovery. The objective of this paper is to compare a model developed by Volvo, Ford, and UMRTI (VFU) through the Advanced Crash Avoidance Technologies (ACAT) Program against a set of real-world departure events. These departure events, collected by Hutchinson and Kennedy, include the vehicle's off road trajectory in 256 road departure events involving passenger vehicles. The VFU-ACAT model was exercised for left side road departures onto the median of a divided highway with a speed limit of 113 kph (70 mph). At low departure angles, the VFU-ACAT model underpredicted the maximum lateral and longitudinal distances when compared to the departure events measured by Hutchinson and Kennedy. Two sets of driver parameters were used to simulate the trajectories, and similar results were seen for the two sets of driver parameters. Vehicles experienced control loss at higher departure angles, particularly in cases modeled with more aggressive driver steering. Maximum lateral and longitudinal distance tended to be overpredicted at high departure angles. This study is part of a larger study that will use the VFU-ACAT driver model to simulate expected benefits of Lane Departure Warning (LDW) and Lane Keeping Assistance (LKA) systems.
Advanced Driver Assistance Systems (ADAS) for collision avoidance/mitigation have already demonstrated their benefit on vehicle safety. Often those systems have an additional functionality for comfort to assist the driver in non-critical driving. The verification of ADAS functionality using different test scenarios is currently investigated in many different projects worldwide. A harmonization of test scenarios and evaluation criteria is not yet accomplished. Often, these test scenarios focus on objective collision avoidance and not on the subjective interaction between driver and vehicle. The present study deals with the development of an experimental validation plan for the systems Automatic Cruise Control (ACC), Lane Departure Warning (LDW) and Lane Keeping Assist (LKA). Standardized driving maneuvers with two or more vehicles equipped with synchronized measurement are performed by professional test drivers. For this purpose selected public roads are used, and the different maneuvers are conducted avoiding critical situations. The evaluation is carried out by several different standardized criteria. Subjective evaluations are correlated with objective results from the measurements and analyzed in a specially prepared evaluation sheet. The present study summarizes the results of this evaluation using different vehicles equipped with relevant ADAS. The results show that there are significant differences in the ADAS behavior which are recognized and evaluated by the drivers. These results are used for future requirements in the vehicle development process. The limitation of the present study is that the spread in evaluation of professional driver has not yet been investigated and correlated to standard driver behavior.
In this paper, switchable Lane Keeping System (LKS) and Active Lane Keeping Assist System (ALKAS) with early/late intervention criteria is proposed and developed. These two systems are commonly based on single track vehicle model and weighted lateral deviation prediction. The main difference is intervention strategy between two systems. Software In the Loop (SIL), Man In the Loop (MIL) verification are fulfilled for both systems till vehicle speed 200kph. For real vehicle verification, only LKS results shall be shown in this paper which shows small maximum lateral deviation also in road transition and vehicle speed variation. Real vehicle verification results for ALKAS shall not be shown because it is more related to steering feeling of driver.
Items per page:
50
1 – 50 of 55