It is the period of the World's history, where the technological progress reached a level that enables the first steps towards the development of vehicles with automated driving capabilities. The swift response from the significant portion of the industry resulted in a race, the final line set at the introduction of vehicles with full automated driving capabilities. Vehicles with automated driving capabilities target making driving safer, more comfortable, and economically more efficient by assisting the driver or by taking responsibilities for different driving tasks. While vehicles with assistance and partial automation capabilities are already in series production, the ultimate goal is in the introduction of vehicles with full automated driving capabilities. Reaching this level of automation will require shifting all responsibilities, including the responsibility for the overall vehicle safety, from the human to the computer-based system responsible for the automated driving functionality (i.e., the Automated Driving System (ADS)). Such a shift makes the ADS highly safe-critical, requiring a safety level comparable to an aircraft system. It is paramount to understand that ensuring such a level of safety is a complex interdisciplinary challenge. Traditional approaches for ensuring safety require the use of fault-tolerance techniques that are unproven when it comes to the automated driving domain. Moreover, existing safety assurance methods (e.g., ISO 26262) suffer from requirements incompleteness in the automated driving context. The use of artificial intelligence-based components in the ADS further complicate the matter due to their non-deterministic behavior. At present, there is no single straightforward solution for these challenges. Instead, the consensus of cross-domain experts is to use a set of complementary safety methods that together are sufficient to ensure the required level of safety. In the context of that, runtime monitors that verify the safe operation of the ADS during execution, are a promising complementary approach for ensuring safety. However, to develop a runtime monitoring solution for ADS, one has to handle a wide range of challenges. On a conceptual level, the complex and opaque technology used in ADS often make researchers ask the question ``how should ADS be verified in order to judge it is operating safely?". Once the initial Runtime Verification (RV) concept is developed, researchers and practitioners have to deal with research and engineering challenges encountered during the realization of the RV approaches into an actual runtime monitoring solution for ADS. These challenges range from, estimating different safety parameters of the runtime monitors, finding solutions for different technical problems, to meeting scalability and efficiency requirements. The focus of this thesis is to propose novel runtime monitoring solutions for verifying the safe operation of ADS. This encompasses (i) defining novel RV approaches explicitly tailored for automated driving, and (ii) developing concepts, methods, and architectures for realizing the RV approaches into an actual runtime monitoring solution for ADS. Contributions to the former include defining two runtime RV approaches, namely the Computer Vision Monitor (CVM) and the Safe Driving Envelope Verification. Contributions to the latter include (i) estimating the sufficient diagnostic test interval of the runtime verification approaches (in particular the CVM), (ii) addressing the out-of-sequence measurement problem in sensor fusion-based ADS, and (iii) developing an architectural solution for improving the scalability and efficiency of the runtime monitoring solution. / RetNet

For people who are blind or have low vision (BLV), physical barriers and negative experiences related to using current transportation options can have negative impacts on quality of life. The emergence of levels 4 – 5 automated driving system-dedicated vehicles (L4+ ADS), which will not require human operators to provide any input into the dynamic driving task, could empower the BLV community by providing an independent means of transportation. Yet, the BLV community has concerns that their needs are not being adequately considered by those currently developing L4+ ADSs, which will result in this technology being inaccessible to populations that it would otherwise greatly benefit. The current study sought to address this gap in the literature by explicitly evaluating the information and interactions that BLV riders will require from L4+ ADS. Specifically, we collected focus group and empirical data across three studies on BLV riders' information and interaction requirements for L4+ ADSs across expected and unexpected driving scenarios as well as pick-up and drop-off tasks (PUDO). Through focus groups with sighted (n = 11) and BLV participants (n = 11; Study 1), we identified similarities and differences between sighted and BLV participants in terms of their user needs for L4+ ADSs across five challenging driving scenarios. Next, we examined BLV participants' (n = 13; Study 2) information requests in real-world settings to better understand BLV riders' needs during a simulated L4+ ADS experience. Our findings show that BLV riders want information that helps with (a) orienting to important objects in the environment during PUDO, (b) determining their location while riding in the ADS, and (c) understanding the ADSs' actions. Finally, we developed an HMI prototype using BLV riders' feedback in Studies 1 and 2 and had BLV participants engage with it during a simulated L4+ ADS trip (n = 12; Study 3). Our results suggest that BLV riders value information about nearby landmarks in familiar and unfamiliar areas, as well as explanations for ADS's actions during ordinary and unexpected scenarios. Additionally, BLV riders need information about required walking distances and presence of tripping hazards in order to select a drop-off location. Taken together, our studies show that BLV riders have specific requirements that L4+ ADS must meet in order for this to be an accessible means of transportation. In light of these findings, we generated 28 guidelines and 44 recommendations that could be used by designers to improve the accessibility of L4+ ADSs for BLV riders. / Doctor of Philosophy / When using current transportation options, individuals who are blind or have low vision (BLV) often encounter physical barriers and negative experiences, which can limit their ability to travel independently and have negative impacts on their overall quality of life. However, future vehicles equipped with levels 4 – 5 automated driving systems (L4+ ADSs) will offer transportation that requires no input from human operators, and thus, could be used as an independent means of transportation for the BLV community. Unfortunately, the BLV community has concerns that their needs are not being adequately considered by those currently developing L4+ ADSs, which will result in this technology being inaccessible to populations that it would otherwise greatly benefit. The current work sought to address this gap in the literature by evaluating the information and interactions that BLV riders will require from L4+ ADS. We conducted three studies to collected data on BLV riders' information and interaction requirements for L4+ ADSs across a variety of driving scenarios as well as tasks relating to being picked up and dropped-off by an L4+ ADS. First, through focus groups with sighted and BLV participants, we identified similarities and differences between sighted and BLV participants' user needs for L4+ ADSs across five challenging driving scenarios. Next, to better understand BLV riders' needs, we had BLV participants indicate when they would desire information during a simulated L4+ ADS ride-hailing experience in real-world settings. Our findings show that BLV riders want information that helps with (a) orienting to important objects in the environment during PUDO, (b) determining their location during their trip, and (c) understanding the reason for the ADS's actions. Finally, using BLV riders' feedback, we developed an HMI prototype and had BLV participants engage with it during a simulated L4+ ADS trip. Our results suggest that BLV riders value information about nearby landmarks in both familiar and unfamiliar areas, as well as explanations for ADS's actions during common (e.g., stopping at a stop sign) and unexpected driving scenarios (e.g., sudden swerve). Additionally, when being dropped off, BLV riders need information about required walking distances and presence of tripping hazards in order to select a desirable drop-off location. Taken together, our studies show that BLV riders have specific requirements that L4+ ADS must meet in order for this to be an accessible means of transportation. In light of these findings, we generated a set of guidelines and recommendations that designers can use to improve the accessibility of L4+ ADSs for BLV riders.

Objectives: Research suggests the general public has a lack of faith in highly automated vehicles (HAV) stems from a lack of system transparency while in motion (e.g., the user not being informed on roadway perception or anticipated responses of the car in certain situations). This problem is particularly prevalent in public transit or ridesharing applications, where HAVs are expected to debut, and when the user has minimal training on, and control over, the vehicle. To improve user trust and their perception of comfort and safety, this study aimed to develop more detailed and tailored human-machine interfaces (HMI) aimed at relying automated vehicle intended actions (i.e., "intentions") and perceptions of the driving environment to the user. Methods: This project developed HMI systems, with a focus on visual and auditory displays, and implemented them into a HAV developed at the Virginia Tech Transportation Institute (VTTI). Volunteer participants were invited to the Smart Roads at VTTI to experience these systems in real-world driving scenarios, especially ones typically found in rideshare or public transit operations. Participant responses and opinions about the HMIs and their perceived levels of comfort, safety, trust, and situational awareness were captured via paper-based surveys administered during experimentation. Results: There was a considerable link found between HMI modality and users' reported levels of comfort, safety, trust, and situational awareness during experimentation. In addition, there were several key behavioral factors that made users more or less likely to feel comfortable in the HAV. Conclusions: Moving forward, it will be necessary for HAVs to provide ample feedback to users in an effort to increase system transparency and understanding. Feedback should consistently and accurately represent the driving landscape and clearly communicate vehicle states to users. / Master of Science / One of the greatest barriers to the entry of highly automated vehicles (HAV) into the market is the lack of user trust in the vehicle. Research has shown that this lack of faith in the system primarily stems from a lack of system transparency while in motion (e.g., the user not being told how the car will react in a certain situation) and not having an effective way to control the vehicle in the event of a system failure. This problem is particularly prevalent in public transit or ridesharing applications, where HAVs are expected to first appear and where the user has less training and control over the vehicle. To improve user trust and perceptions of comfort and safety, this study developed human-machine interface (HMI) systems, focusing on visual and auditory displays, to better relay automated vehicle "intentions" and the perceived driving environment to the user. These HMI systems were then implemented into a HAV developed at the Virginia Tech Transportation Institute (VTTI) and tested with volunteer participants on the Smart Roads.

Objective: In order to introduce automated vehicles on public roads, it is necessary to ensure that these vehicles are safe to operate in traffic. One challenge is to prove that all physically possible variations of situations can be handled safely within the operational design domain of the vehicle. A promising approach to handling the set of possible situations is to identify a manageable number of logical scenarios, which provide an abstraction for object properties and behavior within the situations. These can then be transferred into concrete scenarios defining all parameters necessary to reproduce the situation in different test environments. Methods: This article proposes a framework for defining safety-relevant scenarios based on the potential collision between the subject vehicle and a challenging object, which forces the subject vehicle to depart from its planned course of action to avoid a collision. This allows defining only safety-relevant scenarios, which can directly be related to accident classification. The first criterion for defining a scenario is the area of the subject vehicle with which the object would collide. As a second criterion, 8 different positions around the subject vehicle are considered. To account for other relevant objects in the scenario, factors that influence the challenge for the subject vehicle can be added to the scenario. These are grouped as action constraints, dynamic occlusions, and causal chains. Results: By applying the proposed systematics, a catalog of base scenarios for a vehicle traveling on controlled-access highways has been generated, which can directly be linked to parameters in accident classification. The catalog serves as a basis for scenario classification within the PEGASUS project. Conclusions: Defining a limited number of safety-relevant scenarios helps to realize a systematic safety assurance process for automated vehicles. Scenarios are defined based on the point of the potential collision of a challenging object with the subject vehicle and its initial position. This approach allows defining scenarios for different environments and different driving states of the subject vehicle using the same mechanisms. A next step is the generation of logical scenarios for other driving states of the subject vehicle and for other traffic environments.

The latest trends in autonomous vehicles research gave rise to the needs for specific tools to validate and test such systems. The estimations state that to consider an autonomous vehicle statistically safe, it should drive for thousands of kilometres using traditional validation methods. This process would take a long time. Furthermore, an update in the software, would require to re-run those kilometres. Therefore, the testing must be performed exploiting virtual simulations that should realistically reflect the real world. One way to perfor msuch simulations is to let the vehicle model drive down a road map and control the surrounding traffic. To be effective, spawned traffic should not be generated too far from the target vehicle. The OpenSCENARIO standard offers a feature restricting such traffic within an ellipse centred in the central object (target vehicle). This thesis investigated what technique was more efficient and scalable to detect viable roads within the ellipse to spawn stochastic traffic on. The explored solutions are two: an analytical approach and an adaptation of the AABB tree algorithm. The research started with simple cases and incremented the scenario’s complexity during the development. Through this methodology, each technique’s positive aspects and limits have been highlighted, allowing a comparison to be made. / De senaste trenderna i autonoma fordon har ökat behovet av specifika verktyg för att validera och testa sådana system. För att kunna betrakta ett autonomt fordon som statistiskt säkert, ska enligt uppskattningar autonoma fordon köra tusentals kilometer med traditionella valideringsmetoder. Denna process skulle ta mycket lång tid. Dessutom skulle en uppdatering i mjukvaran kräva att alla dessa tusentals kilometer att körs igen. Därför måste testningen utföras med hjälp av virtuella simuleringar som bör efterlikna den reella världen realistiskt. Ett sätt att genomföra dessa simuleringar är att låta en autonom fordonsmodell köra genom ett vägnät och kontrollera kringliggande trafik. För att vara effektiv, bör kringliggande trafik inte genereras för långt bort från autonoma fordonsmodellen. OpenSCENARIO-standarden innehåller en funktion som begränsar genererad trafik inom en ellips centrerad kring fordonsmodellen. Detta examensarbete undersökte vilka tekniker som är mest effektiva och skalbara för att detektera relevanta vägar inom ellipsen att generera stokastisk trafik på. De två lösningar som studerades var: en analytisk och en numerisk som använde sig av AABB-träd-algoritmen. Utförandet började med simpla fall som successivt ökade till mer avancerade scenarion. Genom denna metodik blev varje tekniks positiva aspekter samt begränsningar belysta och jämförbara.

<p dir="ltr">Automated Driving Systems (ADS) have evolved significantly over the past decade. With conditionally automated driving systems still requiring constant driver supervision and human intervention upon system request, a driver’s readiness to take over from an ADS has significant safety implications. Research suggests that drivers using ADS are more likely to engage in non-driving-related tasks (NDRTs), and this engagement can deteriorate takeover performance. However, different NDRTs can involve engagement of physical, visual and/or cognitive resources, which all can affect the takeover process in different ways. The potential interaction effects among these factors may be the cause of mixed empirical findings regarding the influence of NDRT engagement on takeover readiness and performance. Additionally, with more advanced ADS, takeover scenarios are likely to be less urgent. Yet, the ways in which drivers behave in response to a takeover request to intervene during such less urgent scenarios while engaged in NDRTs is still not well understood.</p><p dir="ltr">The purpose of this dissertation is to provide a better understanding of drivers’ response behavior during a conditionally automated vehicle takeover process by analyzing drivers’ motor, visual, and cognitive readiness in response to a takeover request (TOR). The work was completed in two phases. The first phase focused on the effects of pre-takeover visual engagement on takeover readiness in urgent situations. Two experiments were conducted as part of this first phase. Particularly, Study 1 investigated drivers’ post-TOR visual attention allocation and cognitive readiness after continuous visual NDRT engagement before a TOR. Study 2 examined drivers’ pre-TOR visual attention allocation and takeover performance both during and after voluntary engagement with visual NDRTs. The second phase used a non-urgent takeover scenario to investigate drivers’ takeover behavior and visual attention allocation when prioritizing the engagement of visual-manual NDRTs that differed in terms of cognitive engagement levels.</p><p dir="ltr">Study 1 required continuous visual attention in NDRTs and manipulated only the location of visual attention before an auditory TOR. Dependent measures included duration, location, and directness eye-tracking measures after the TOR, as well as freeze-probe cognitive readiness scores. Overall, delayed visual attention re-allocation in the driving scene, less dispersed gaze patterns, and worse perception and comprehension of road hazards were associated with off-road visual NDRT engagement. In addition, no significant benefit of enforcing on-road visual attention before the TOR, compared to the baseline condition without NDRT requirements, were found. These findings highlight the need to investigate the effects of more naturalistic NDRT engagement on takeover attention reallocation and takeover performance.</p><p dir="ltr">Study 2 complemented Study 1 by allowing voluntary switching of visual attention between the NDRT and the driving scene prior to the TOR, with the driving task being a priority. In addition, Study 2 investigated drivers’ takeover quality and understanding of the takeover scene using the appropriateness of their takeover decisions. Dependent measures were pre- and post-takeover eye-tracking measures, aligning to those used in Study 1, in addition to motor response measures, longitudinal and lateral vehicle control measures, and decisions made in response to a road obstacle. Overall, the driver’s post-TOR behaviors were not significantly affected by NDRT conditions, but visual NDRT-induced differences in gaze distribution were associated with the appropriateness of takeover decisions.</p><p dir="ltr">Finally, Study 3 used knowledge from prior studies to isolate the effects of different levels of cognitive engagement in real-world visual-manual NDRTs. The purpose was to investigate the effects of cognitive engagement on drivers’ visual attention allocation before and during the takeover, as well as on takeover performance in non-urgent takeover scenarios, where NDRT engagement was a priority. Dependent measures included eye-tracking measures, takeover response time, and vehicle control measures, used in prior studies. In summary, engagement in NDRTs with higher levels of cognitive engagement resulted in significant differences in pre-TOR visual attention allocation and less stable takeover maneuvers.</p><p dir="ltr">The findings from this work contribute to a better understanding of the effects of different components of NDRT engagement on takeover performance in conditionally automated driving systems. Ultimately, this work can contribute to improving the design of next-generation human-machine interfaces in surface transportation, including driver monitoring systems and in-vehicle displays, that promote safer human-automation integration in future ADS.</p>

Advanced Vehicle Systems promise improved safety and comfort for drivers. Steady advancements in technology are resulting in increasing levels of vehicle automation capabilities, furthering safety benefits. In fact, some of these vehicle automation systems are already deployed and available, but with promised benefits, such systems can potentially change driving behaviors. There is evidence that drivers have increased secondary task engagements while driving with automated vehicle systems, but there is a need for a clearer scientific understanding of any potential correlations between the use of automated vehicle systems and potentially negative driver behaviors. Therefore, this thesis aims to understand the state of knowledge on automated vehicle systems and their possible impact on drivers’ distraction behaviors. I have conducted two systematic literature reviews to examine this question. This thesis reports these reviews and examines the effects of secondary task engagement on driving behaviors such as take-over times, visual attention, trust, and workload, and discusses the implications on driver safety.