The goal of this thesis is to implement the control and design of an intelligent vehicle based on an embedded DSP platform (eZdspTM F2812). The overall system including steering wheel AC serve motor, brake actuator, throttle driving circuit and the sensors is equipped in a golf car as a platform. Otherwise, digital image processing technology is used to realize the autonomous driving system which can achieve multi-mode of lane-keeping, lane-change and obstacle-avoidance. In the lane-keeping control, the road information can be provided by the vision system. According to the offset and displacement of angle as input signal, a fuzzy controller is used to compute the desired steering wheel angle and let the golf car can cross road safely. In lane-change, a smooth trajectory can be generated by IMU, IMU is used to collect the information data of yaw rate and yaw angle when human-driving. That makes autonomous driving system become more humanlike and to achieve an open-loop lane-change maneuver. In obstacle-avoidance, we use a laser range scanner to detect the distance of a front obstacle. When the distance is lower than safety distance, double lane-change will be activated to avoid the front obstacle. The overall system has been examined on NSYSU campus roads.

Road departure crashes are among the deadliest crash modes in the U.S. each year. In response, automakers have been developing lane departure active safety systems to alert drivers to impending departures. These lane departure warning (LDW) and lane departure prevention (LDP) systems have great potential to reduce the frequency and mitigate the severity of serious lane and road departure crashes. The objective of this thesis was to characterize lane and road departures to better understand the effect of systems such as LDW and LDP on single vehicle road departure crashes. The research includes the following: 1) a characterization of lane departures through analysis of normal lane keeping behavior, 2) a characterization of road departure crashes through the development and validation of a real-world crash database of road departures (NCHRP 17-43 Lite), and 3) develop enhancements to the Virginia Tech LDW U.S. fleetwide benefits model. Normal lane keeping behavior was found to vary with road characteristics such as lane width and road curvature. Consideration of the dynamic driving behaviors observed in the naturalistic driving study (NDS) data is important to avoid LDW false alarms and driver annoyance. Departure characteristics computed in normal driving were much less severe than the departure parameters measured in real-world road departure crashes. The real-world crash data collected in NCHRP 17-43 Lite database was essential in developing enhancements to the existing Virginia Tech LDW fleetwide benefits model. Replacement of regression model predictions with measured crash data and improvement of the injury criteria resulted in an 11-16% effectiveness for road departure crashes, and an 11-15% reduction in seriously injured drivers. / Master of Science / Road departure crashes account for nearly one-third of the roughly 30,000 automobile traffic fatalities in the U.S. each year. Lane departure warning (LDW) and lane departure prevention (LDP) systems are two safety systems developed to reduce the large number of fatalities resulting from road departures. The safety systems warn drivers if the vehicle begins to drift out of the intended lane of travel, and automatically steer the vehicle back into the lane of travel if it continues to drift. While LDW and LDP systems have potential to lower the number of fatal lane and road departure crashes, the technology is not yet a standard feature in production vehicles. There has been a lower than expected acceptance rate, and real-world benefits of the systems have not been published. The research objective for this thesis was to characterize lane and road departures to investigate the effect of these safety systems on road departure crashes. The first section of this thesis analyzed large amounts of time series data recorded from people in normal driving scenarios to model lane keeping behavior in non-crash, drift out of lane departures. We found driving behavior varied with road characteristics such as lane width and road curvature. These dynamic driving behaviors may lead to LDW false alarms and contribute to driver annoyance with the systems. The second portion of this research involved the development and validation of a real-world road departure crash database. The database included key departure parameters such as angle, speed, and road curvature. These parameters were used in the third section of the thesis to enhance the Virginia Tech LDW U.S. fleetwide benefits model, which is a mathematical trajectory simulation model that determines whether or not these road departure crashes could have been prevented if every vehicle in the U.S. was equipped with LDW. We found an effectiveness of 11-16% prevention for road departure crashes, and an 11-15% reduction in serious driver injury.

<p>Many traffic accidents are caused by driver unawareness. This includes fatigue, drowsiness and distraction. In this thesis two systems are described that could be used to decrease the number of accidents. In the first part of this thesis a system using long-term information to warn drivers suffering from fatigue is developed. Three different versions with different criteria are evaluated. The systems are shown to handle more then 60% of the cases correctly.</p><p>The second part of this thesis examines the possibilities of developing a warning system based on the predicted time-to-lane crossing, TLC. A basic TLC model is implemented and evaluated. For short time periods before lane crossing this may offer adequate accuracy. However the accuracy is not good enough for the model to be used in a TLC based warning system to warn the driver of imminent lane departure.</p>

Many traffic accidents are caused by driver unawareness. This includes fatigue, drowsiness and distraction. In this thesis two systems are described that could be used to decrease the number of accidents. In the first part of this thesis a system using long-term information to warn drivers suffering from fatigue is developed. Three different versions with different criteria are evaluated. The systems are shown to handle more then 60% of the cases correctly. The second part of this thesis examines the possibilities of developing a warning system based on the predicted time-to-lane crossing, TLC. A basic TLC model is implemented and evaluated. For short time periods before lane crossing this may offer adequate accuracy. However the accuracy is not good enough for the model to be used in a TLC based warning system to warn the driver of imminent lane departure.

The Indiana Department of Transportation (INDOT) manages and maintains over 3,000 miles of interstates across the state. Assessing lane marking quality is an important part of agency asset tracking and typically occurs annually. The current process requires agency staff to travel the road and collect representative measurements. This is quite challenging for high volume multi-lane facilities. Furthermore, it does not scale well to the additional 5,200 centerline miles of non-interstate routes. <div><br></div><div>Modern vehicles now have technology on them called “Lane Keep Assist” or LKA, that monitor lane markings and notify the driver if they are deviating from the lane. This thesis evaluates the feasibility of monitoring when the LKA systems can and cannot detect lane markings as an alternative to traditional pavement marking asset management techniques. This information could also provide guidance on what corridors are prepared for level 3 autonomous vehicle travel and which locations need additional attention. </div><div><br></div><div>In this study, a 2019 Subaru Legacy with LKA technology was utilized to detect pavement markings in both directions along Interstates I-64, I-65, I-69, I-70, I-74, I90, I-94 and I-465 in Indiana during the summer of 2020. The data was collected in the right most lane for all interstates except for work zones that required temporary lane changes. The data was collected utilizing two go-pro cameras, one facing the dashboard collecting LKA information and one facing the roadway collecting photos of the user’s experience. Images were taken at 0.5 second frequency and were GPS tagged. Data collection occurred on over 2,500 miles and approximately 280,000 images were analyzed. The data provided outputs of: No Data, Excluded, Both Lanes Not Detected, Right Lane Not Detected, Left Lane Not Detected, and Both Lanes Detected. </div><div><br></div><div>The data was processed and analyzed to create spatial plots signifying locations where markings were detectable and locations where markings were undetected. Overall, across 2,500 miles of travel (right lane only), 77.6% of the pavement markings were classified as both detected. The study found</div><div><br></div><div>• 2.6% the lane miles were not detected on both the left and right side </div><div>• 5.2% the lane miles were not detected on the left side </div><div>• 2.0% the lane miles were not detected on the right side 8 </div><div><br></div><div>Lane changes, inclement weather, and congestion caused 12.5% of the right travel lane miles to be excluded. The methodology utilized in this study provides an opportunity to complement the current methods of evaluating pavement marking quality by transportation agencies. </div><div><br></div><div>The thesis concludes by recommending large scale harvesting of LKA from a variety of vendors so that complete lane coverage during all weather and light conditions can be collected so agencies have an accurate assessment of how their pavement markings perform with modern LKA technology. Not only will this assist in identifying areas in need of pavement marking maintenance, but it will also provide a framework for agencies and vehicle OEM’s to initiate dialog on best practices for marking lines and exchanging information.</div>

Indiana University-Purdue University Indianapolis (IUPUI) / In the past decades, the rapid innovation of technology has greatly affected the automotive industry. However, every innovation has always been paired with safety risks that needs to be quickly addressed. This is where Petri nets (PNs) have come into the picture and have been used to model complex systems for different purposes, such as production management, traffic flow estimation and the introduction of new car features collectively known as, Adaptive Driver Assistance Systems (ADAS). Since most of these systems include both discrete and continuous dynamics, the Hybrid Petri net (HPN) model is an essential tool to model these. The objective of this thesis is to develop, analyze and simulate a lane keeping support system using an HPN model. Chapter 1 includes a brief summary of the specific ADAS used, lane departure warning and lane keeping assist systems and then related work on PNs is mentioned. Chapter 2 provides a background on Petri nets. In chapter 3, we develop a discrete PN model first, then we integrate continuous dynamics to extend it to a HPN model that combines the functionalities of the two independent ADAS systems. Several scenarios are introduced to explain the expected model behavior. Chapter 4 presents the analysis and simulation results obtained on the final model. Chapter 5 provides a summary for the work done and discusses future work.

The lane-keeping system in autonomous vehicles (AV) or even as a part of the advanced driving assistant system (ADAS) is known as one of the primary options of AVs and ADAS. The developed lane-keeping systems work on either computer vision or deep learning algorithms for their lane detection section. However, even the strongest image processing units or the robust deep learning algorithms for lane detection have inaccuracies during lane detection under certain conditions. The source of these inaccuracies could be rainy or foggy weather, high contrast shades of buildings and objects on-street, or faded lines. Since the lane detection unit of these systems is responsible for controlling the steering, even a momentary loss of lane detection accuracy could result in an accident or failure. As mentioned, different lane detection algorithms have been presented based on computer vision and deep learning during the last few years, and each one has pros and cons. Each model may have a better performance in some situations and fail in others. For example, deep learning-based methods are vulnerable to new samples. In this research, multiple models of lane detection are evaluated and used together to implement a robust lane detection algorithm. The purpose of this research is to develop an estimator-based Multiple Model Adaptive Estimation (MMAE) algorithm on the lane-keeping system to improve the robustness of the lane detection system. To verify the performance of the implemented algorithm, the AirSim simulation environment was used. The test simulation vehicle was equipped with one front camera and one back camera used to implement the proposed algorithm. The front camera images are used for detecting the lane and the offset of the vehicle and center point of the lane. The rear camera, which offered better performance in lane detection, was used as an estimator for calculating the uncertainty of each model. The simulation results showed that combining two implemented models with MMAE performed robustly even in those case studies where one of the models failed. The proposed algorithm was able to detect the failures of either of the models and then switch to another good working model to improve the robustness of the lane detection system. However, the proposed algorithm had some limitations; it can be improved by replacing PID controller with an MPC controller in future studies. In addition, in the presented algorithm, two computer vision-based algorithms were used; however, adding a deep learning-based model could improve the performance of the proposed MMAE. To have a robust deep learning-based model, it is suggested to train the network based on AirSim output images. Otherwise, the network will not work accurately due to the differences in the camera's location, camera configuration, colors, and contrast.

Indiana University-Purdue University Indianapolis (IUPUI) / The lane-keeping system in autonomous vehicles (AV) or even as a part of the advanced driving assistant system (ADAS) is known as one of the primary options of AVs and ADAS. The developed lane-keeping systems work on either computer vision or deep learning algorithms for their lane detection section. However, even the strongest image processing units or the robust deep learning algorithms for lane detection have inaccuracies during lane detection under certain conditions. The source of these inaccuracies could be rainy or foggy weather, high contrast shades of buildings and objects on-street, or faded lines. Since the lane detection unit of these systems is responsible for controlling the steering, even a momentary loss of lane detection accuracy could result in an accident or failure. As mentioned, different lane detection algorithms have been presented based on computer vision and deep learning during the last few years, and each one has pros and cons. Each model may have a better performance in some situations and fail in others. For example, deep learning-based methods are vulnerable to new samples. In this research, multiple models of lane detection are evaluated and used together to implement a robust lane detection algorithm. The purpose of this research is to develop an estimator-based Multiple Model Adaptive Estimation (MMAE) algorithm on the lane-keeping system to improve the robustness of the lane detection system. To verify the performance of the implemented algorithm, the AirSim simulation environment was used. The test simulation vehicle was equipped with one front camera and one back camera used to implement the proposed algorithm. The front camera images are used for detecting the lane and the offset of the vehicle and center point of the lane. The rear camera, which offered better performance in lane detection, was used as an estimator for calculating the uncertainty of each model. The simulation results showed that combining two implemented models with MMAE performed robustly even in those case studies where one of the models failed. The proposed algorithm was able to detect the failures of either of the models and then switch to another good working model to improve the robustness of the lane detection system. However, the proposed algorithm had some limitations; it can be improved by replacing PID controller with an MPC controller in future studies. In addition, in the presented algorithm, two computer vision-based algorithms were used; however, adding a deep learning-based model could improve the performance of the proposed MMAE. To have a robust deep learning-based model, it is suggested to train the network based on AirSim output images. Otherwise, the network will not work accurately due to the differences in the camera's location, camera configuration, colors, and contrast.

Die Arbeit befasst sich mit der Entwicklung automatischer Lenkungen, die die von Schienenfahrzeugen bekannte Spurtreue auf n-gliedrige, mehrachsgelenkte Straßenfahrzeuge übertragen. Spurtreu bedeutet folglich, dass die Lenkachsmittelpunkte keinen seitlichen Versatz zueinander aufweisen. Dafür wird ein modellbasiertes automatisches Lenkverfahren systematisch konzipiert, entworfen und erprobt, das sowohl eine vollautomatische Spurführung als auch eine halbautomatische Nachführung erlaubt. Die modellbasierten automatischen Lenkungen unterliegen keinen praktisch relevanten Einschränkungen. Das wird durch die Verwendung einer Steuerungsstruktur mit zwei Freiheitsgraden erreicht, die aus einer modellbasierten Vorsteuerung und einem Rückführregler besteht. In der Vorsteuerung werden die Lenkwinkel aller Achsen berechnet, mit denen der Sollweg theoretisch spurtreu befahren wird. Durch den Einsatz eines speziell angepassten, modularen Mehrkörpermodells gelingt diese Berechnung allgemein für eine Klasse n-gliedriger Fahrzeuge. Zum Ausgleich von nicht vermeidbaren Modellunbestimmtheiten und nicht gemessenen Störungen werden ein nichtlinearer Mehrgrößenregler sowie achs-individuelle lineare Eingrößenregler entworfen und miteinander verglichen. Simulationen und Fahrversuche zeigen, dass das entwickelte Verfahren in einem weiten Geschwindigkeitsbereich robust gegenüber typischen Einflussgrößen wie Fahrbahn- und Beladungszustand ist.

Objective: The Vision Zero initiative pursues the goal of eliminating all traffic fatalities and severe injuries. Today’s advanced driver assistance systems (ADAS) are an important part of the strategy toward Vision Zero. In Germany in 2018 more than 26,000 people were killed or severely injured by traffic accidents on motorways and rural roads due to road accidents. Focusing on collision avoidance, a simulative evaluation can be the key to estimating the performance of state-of-the-art ADAS and identifying resulting potentials for system improvements and future systems. This project deals with the effectiveness assessment of a combination of ADAS for longitudinal and lateral intervention based on German accident data. Considered systems are adaptive cruise control (ACC), autonomous emergency braking (AEB), and lane keeping support (LKS). Methods: As an approach for benefit estimation of ADAS, the method of prospective effectiveness assessment is applied. Using the software rateEFFECT, a closed-loop simulation is performed on accident scenario data from the German In-Depth Accident Study (GIDAS) precrash matrix (PCM). To enable projection of results, the simulative assessment is amended with detailed single case studies of all treated cases without PCM data. Results: Three categories among today’s accidents on German rural roads and motorways are reported in this study: Green, grey, and white spots. Green spots identify accidents that can be avoided by state-of-the-art ADAS ACC, AEB, and LKS. Grey spots contain scenarios that require minor system modifications, such as reducing the activation speed or increasing the steering torque. Scenarios in the white category cannot be addressed by state-of-the-art ADAS. Thus, which situations demand future systems are shown. The proportions of green, grey, and white spots are determined related to the considered data set and projected to the entire GIDAS. Conclusions: This article describes a systematic approach for assessing the effectiveness of ADAS using GIDAS PCM data to be able to project results to Germany. The closed-loop simulation run in rateEFFECT covers ACC, AEB, and LKS as well as relevant sensors for environment recognition and actuators for longitudinal and lateral vehicle control. Identification of green spots evaluates safety benefits of state-of-the-art level 0–2 functions as a baseline for further system improvements to address grey spots. Knowing which accidents could be avoided by standard ADAS helps focus the evolution of future driving functions on white spots and thus aim for Vision Zero.