January 21, 2020 Feature

Automated and Assistive Vehicle Technology: Opportunities and Challenges

By David Cades, Carmine Senatore, John L. Campbell, Ryan Harrington, and Daniel Wood

Image from Getty Images

In the past two decades, there has been a sharp increase in the development and availability of technology related to automated vehicles and advanced driver-assistance systems (ADAS), generally aimed at improving safety and mobility for users of the transportation system. With the advent of increasingly automated driver-assistance features, the industry has been presented with new challenges as these innovative technologies are deployed and consumers are exposed to technologies with varying capabilities and limitations.

Levels of Driving Automation

SAE International (formerly the Society of Automotive Engineers) has provided a framework to discuss advanced and automated vehicle technology and facilitate understanding of this technology in its J3016 standard, Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driving Systems. This standard describes six levels of automation, ranging from level 0 (momentary assistance) to level 5 (full automation). As the levels increase from 0 to 5, the amount of automation increases, and the role and control of the driver decrease. The levels are further grouped such that levels 0–2 are considered driver-support features, whereby the driver must be attentive and engaged in driving at all times even though the vehicle’s automated features may intervene or support the driving task; and levels 3–5 specify various conditions under which the vehicle is automated, and the driver does not have a role in the driving task when these features are engaged.

Features of levels. Specifically, as can be seen in Figure 1, level 0 includes features like automatic emergency braking (AEB) and blind spot detection (BSD), while lane centering assist (LCA) and adaptive cruise control (ACC) can be ascribed (individually) to level 1.

The most advanced technologies currently on the general market are level 2; an example is the concurrent operation of LCA and ACC, which, together, maintain the vehicle’s position on the roadway and relative to other vehicles, under certain circumstances. However, to ensure the safe operation of the vehicle, the driver must continue to monitor the driving environment.

Level 3 systems, which are described as full automation under specific conditions, are not available on the mainstream market yet; however, several systems have been announced for upcoming market deployment. One of the key considerations in a level 3 system is that the system and the driver have to hand off control of the vehicle to each other at particular times. Therefore, even when the vehicle automation is enabled, a driver must remain attentive to be able to promptly regain awareness of the environment and take control of the driving task.

Moving up the scale, levels 4 and 5 represent full automation of the driving task and do not require the driver to take over the driving task. In practical terms, the primary difference in these upper levels lies in the fact that the vehicle occupant does not have to remain engaged in the driving task. Level 4 provides for full automation only under certain driving conditions or in certain specified geographic areas, and level 5 provides for full automation under all driving conditions. Levels 4 and 5 systems currently are being tested, some on public roadways, but time to market for these systems is still uncertain. See Figure 1 on page 19, which is a graphic depicting the six levels of driving automation.

Incident investigations: technology concerns. Most of the automated technology at issue in current incident investigations of equipped (or nonequipped) vehicles would be categorized as ADAS (i.e., SAE levels 0–2). In considering incidents involving these technologies, a few technology-specific questions become immediately relevant:

  1. Should the vehicle at issue have been equipped with a given advanced driver-assistance system?
  2. Was a better technology available?
  3. Did the technology function as intended?
  4. How did the design and specific parameters of the technology effectively interact with the user?
  5. What information was provided to the user with respect to the capabilities and limitations of the technology?

The technical and scientific concepts related to these questions highlight opportunities and challenges for incident investigators, regulators, and insurers, as well as system designers as they work on the development and deployment of these technologies while protecting consumers.

Technology Development and Deployment

ADAS. While levels 4 and 5 vehicles are still in the development phase when it comes to direct consumer access, ADAS technologies (levels 0–2) have been present and growing in the U.S. fleet since the early 2000s (SAE level 3 systems are in the process of being introduced on select vehicles and are available in limited numbers in certain international markets).

ADAS include a variety of systems, which range from providing collision warnings to drivers to actively assisting drivers in the driving task through, for example, LCA or AEB. Individual ADAS generally are designed to support one specific aspect of the driving task and in general cannot (1) completely substitute for the driver, (2) operate in all environmental conditions, or (3) be 100 percent effective 100 percent of the time.

In 2006, Acura’s Collision Mitigation Braking System (CMBS) and Mercedes-Benz’s Brake Assist BAS Plus on its flagship S-Class model were the first radar-based systems capable of providing automatic braking. In the 2006–2008 time frame, other vehicle manufacturers followed suit with similar systems on their higher-end vehicles.

S curve. While more and more vehicles are being equipped with ADAS, such technology has not been deployed yet on every vehicle being produced and sold. This is not surprising as new technology deployment typically follows an S curve of market penetration, as seen in Figure 2 below. Historically, the first 10 to 15 years after a technology’s introduction are characterized by a low-volume deployment rate (the flat part of the curve), followed by an increase in the deployment rate as the technology matures, and often taking another five to 10 years to reach the high-volume deployment rate on the top, or right, end of the S.

The shape of the S curve is defined by, among other things, the following:

  • Capital investment and engineering resources,
  • Technological breakthroughs and intellectual property (IP) considerations,
  • Maturity of manufacturing processes,
  • Design, production, and durability validation testing,
  • Vehicle refresh and redesign cycles,
  • Supplier capacity,
  • Consumer acceptance and affordability, and
  • Voluntary or mandatory standards.

NHTSA and deployment rates. When the National Highway Traffic Safety Administration (NHTSA) promulgates rulemakings (e.g., Corporate Average Fuel Economy (CAFE) standards), it expends considerable effort understanding and accounting for the deployment rate of a technology in its analyses.

For the CAFE rulemakings, emerging technologies are assumed to penetrate the market only in the single-digit percentage growth rate range per year; and even mature technologies, which are in the maximum growth rate regime of the S curve, are assumed to penetrate the market at growth rates of only 10 percent to 20 percent per year. The time horizon for the entire process of a technology becoming standard equipment on all vehicles produced can be 15 to 30 years. Furthermore, when considering the entire fleet, historical data suggests that it can, and does, take many decades for a technology to be present on nearly every vehicle on the road.

The 2017 announcement between the NHTSA, the Insurance Institute for Highway Safety (IIHS), and 20 vehicle manufacturers regarding those manufacturers’ voluntary commitment to introduce AEB as standard equipment on most vehicles by model year 2022 seems to acknowledge the NHTSA’s understanding of the current state of AEB (i.e., still an evolving, not yet fully mature technology) and the time required to achieve widespread deployment of the emerging AEB technology.1 Even with this voluntary commitment to equip new vehicles with AEB, and with the impending transition into the max growth rate phase of the S curve, it still will take decades past 2022 before a substantial portion of the entire U.S. fleet of vehicles on the road will be AEB-equipped. Higher levels of automated vehicles, on the other hand, are still in the lower left corner of the S curve (in the low-volume deployment region) and will require more time to enter the max growth rate phase.

User Trust and Acceptance

As vehicle manufacturers, suppliers, and technology developers bring ADAS and automated technologies to market, in addition to working within the frameworks provided by federal, state, and local authorities, care must be taken to consider how consumers will accept and use these systems. Regardless of how the technology functions, for any of these systems to have the desired effects that they often advertise, the ADAS-equipped vehicles must get into the hands of drivers, and the systems must be used by the drivers. For this to occur, consumers must be able to afford, perceive the benefits of, and ultimately use ADAS technologies.

Affordability. Affordability is one of the forces that drives newer technologies’ debut in higher-end luxury makes and models because these vehicles and their prospective owners are more likely to bear the cost of new and expensive technology. Referring to the S curve mentioned above, an increase in market penetration of ADAS technology typically occurs as the technology enters a mature stage with associated lower production costs.

Trust. As these vehicles get into the hands of more users and the technology continues to mature and proliferate, the success of these systems’ emergence also will be influenced by how the users trust the technology. Research on trust in automation has shown that an unreliable system can cause the user to lose trust and will be underutilized—and, thus, not effective due to nonuse.2 At worst, the user will disable the system entirely.3 Conversely, when a system performs reliably and is accepted by the user, trust (and use) can be facilitated. One study comparing drivers’ reactions to adaptive cruise control systems with varying levels of reliability showed that trust grew over time for drivers exposed to a 100 percent reliable system.4 With the knowledge that trust of the technology is integral, and that even infrequent violations of that trust can have a lasting effect on users’ feelings toward automated systems,5 consideration of these issues will go a long way to ensuring that ADAS will be integrated successfully into the automotive marketplace and become common equipment on larger portions of the automotive fleet.

If higher reliability fosters more trust among users, then a key challenge for ADAS development is the accuracy—and perceived accuracy—of hazard alerts that are provided by the system. False alarms or alerts are those provided by the ADAS device that are triggered in the absence of an actual hazard. A false alarm occurs, for example, when the system provides a forward collision alert in the absence of cars or other obstacles in front of the driver’s vehicle or when the alert references an out-of-path vehicle or roadside object such as a guardrail or a sign that is not a hazard to the vehicle or the vehicle’s intended path. Even a small number of false alarms can turn into a nuisance for the operator. Nuisance alerts include a subjective component; they can be triggered by an actual hazard but viewed as inappropriate or unnecessary due to the manner in which they are delivered, e.g., frequency, timing, intensity, or modality.6 Whether or not the nuisance stems from false alarms or alerts that are deemed inappropriate by the user, the achievement of nuisance status will lead to higher rates of disuse and mistrust.

The “truth table” (Table 1) on this page summarizes performance outcomes from alerts provided by ADAS devices.

Usage. Though the literature shows mixed driver responses to false alarms and nuisance alerts, either of these can affect driver behavior, performance, and acceptance of ADAS. For example, one study showed that false alerts resulted in longer brake reaction times immediately following a previous false or missing alert, and that drivers tended to brake for false alarms when they were timed to provide a relatively longer time-to-collision.7 False alerts or alarms therefore can induce carryover effects from previous experiences such that experience with an alert can change behaviors and responses to subsequent alerts.

A field operational test of vehicles equipped with multiple collision-warning technologies (e.g., forward collision, curve speed warning, lane departure warning, and blind spot warning) found a nuisance alarm rate of 0.83 per 100 miles of driving across all warning types.8 Despite this rate of nuisance alarms, 72 percent of all drivers were still interested in having the integrated system in their cars, versus 25 percent who were not interested. However, although many drivers found the nuisance rate to be tolerable, several drivers reported that the false alarms caused them to distrust the system and begin to ignore alerts.9 While ignoring false alarms is appropriate, the danger is that drivers may ignore true positive alerts and not initiate an appropriate response in a hazardous situation that requires a response.

Driver acceptance data obtained from an independent evaluation of this same field operational test was similarly nuanced. One study reported that while false/nuisance alerts were the system characteristic liked least by 50 percent of the drivers, 81 percent still found the warnings helpful, and 82 percent were satisfied with the system overall.10 These findings suggest, in general, that drivers find many of these systems useful but that the specific user experience and design of any individual advanced driver-assistance system will necessarily affect its utility and adoption.

Drivers also may respond differently to false alarms as compared to nuisance alerts. One study found that nuisance alerts were associated with greater compliance to the alert in critical situations, while false alarms were associated with less compliance.11 Overall, false alarms and nuisance alerts can influence drivers’ perception of the reliability of the ADAS device and change their degree of trust in the system;12 impact their responses to alerts; reduce system effectiveness if “true” alerts are ignored; and, finally, as mentioned above, affect the overall efficacy of ADAS technology.

Consumer Education

Another aspect of the driver-ADAS interaction that will be critical to evaluate in order to understand the potential effectiveness of a given technology is the information available with respect to the systems’ capabilities and limitations and how that information is communicated to the driver. Specifically, consumers’ understanding of what the technology can do, and what it cannot do, will directly affect proper use and potential misuse. For example, certain forward collision mitigation systems are programmed to identify stopped cars in the vehicle’s path of travel but may not alert drivers to a different object or shape in their path. Similarly, systems that rely on video processing to identify objects, roadway features, and hazards may not function as well in situations with limited optical ability (e.g., darkness, fog, rain, snow, etc.).

Lack of understanding. That users have difficulty understanding the capabilities and limitations of new technology is not new. In fact, in the early years of anti-lock braking systems (ABS), nearly 50 percent of complaints about ABS were related to occurrences in which the system was functioning as intended. The complaints were due to the users not understanding how the system worked.13

Similar trends have been observed with respect to ADAS and autonomous vehicle technologies. Specifically, one study revealed that less than 80 percent of respondents believed that a vehicle marketed as “Fully Autonomous” would be able to turn corners.14 The study also showed that only about 50 percent of respondents believed that a vehicle marketed with “Driver Assistance” would notify the driver when the driver was needed.15 When investigating a specific incident, it will be important to assess what the driver knew and understood, or could have known and understood, with respect to the available technology and how that may have influenced the driver’s behavior.

Even if a driver is (1) operating a vehicle with a properly functioning ADAS technology and (2) encounters a hazard scenario within the system’s capabilities and operational envelope, it is still possible that the driver will not comprehend or effectively respond to an evolving hazard or a system-generated alert. In order to understand either a situation as it unfolds or a hazard alert and then respond in an appropriate and timely fashion, drivers must develop and maintain a functionally accurate understanding of how the ADAS technology operates. This includes the user’s knowledge of the technology’s purpose, how it works, and how it is likely to work in the future.16 While drivers generally have a good understanding of how common vehicle features operate (such as cruise control), their understanding will likely be weaker, incorrect, or nonexistent for early implementations of vehicles with ADAS technology. Given the many ways in which drivers can interact with such technologies, a challenge faced by ADAS developers is how to support the formation of a functionally accurate understanding of these vehicles.

Greater automation: higher stakes. This issue becomes even more critical with higher levels of vehicle automation, as the human driver’s role shifts toward having less responsibility for monitoring the external environment and actively controlling the vehicle and more responsibility for supervising the automation. In such situations, a lack of understanding of how the system operates can lead to incorrect assumptions about the system’s abilities, confuse the driver, and even contribute to crashes. For example, suppose a vehicle equipped with both ACC and LCA systems is operating on a two-lane highway (i.e., both longitudinal and lateral control is being accomplished by the vehicle). The vehicle approaches a work zone in which the highway quickly narrows to one lane. Without a clear understanding of how the vehicle will behave in such a situation, the driver might (1) assume that the system can properly sense the situation and handle the lane shift—and face the possibility of a crash if it does not; or (2) seize control of the vehicle and initiate a manual lane shift without having the time to check and see if any vehicles are nearby. In both situations, a mismatch between the system’s capabilities and the driver’s understanding of those capabilities creates a potentially unsafe situation.17

Expectations-operation match. In general, greater levels of understanding of what the system can do and how it will operate in different situations are achieved when the driver’s expectations of and experience with the technology are aligned with its actual operation and capabilities. Such operational consistency is crucial in the development of an accurate understanding of how advanced vehicle technology works. In the case of ACC technology, one study notes that the utility of the automation is a function of not only the driver’s understanding of the vehicle-environment interaction but also the driver’s understanding of the automation.18 When the ACC system behaves in a manner consistent with that of a reasonable driver, the driver’s understanding of the system will be supported, and the driver will be more likely to intervene when needed.19

Benefits of education. There are many benefits associated with helping drivers develop an accurate understanding of how ADAS technologies operate. For example, research in a variety of domains has identified that having a functionally accurate understanding of an automated system is a central aspect in reaching a desired level of expertise with the system,20 improving users’ level of trust in the system,21 and improving users’ ability to identify errors in the system’s operation.22 As detailed above, all of these traits generally lead to more use, more proper use, and better performance with automated technology and would do the same for ADAS.

Incident Investigations and Advanced Vehicle Technologies

In the context of the issues and scientific principles mentioned above, it is clear that the investigation into (and corresponding evaluation of) the potential effectiveness of an ADAS-equipped or ADAS-nonequipped vehicle with respect to an individual incident (or series of incidents) is not as simple as a binary presence-or-absence discussion. This section discusses some of the aspects that will be critical for investigators to consider when dealing with incidents either involving ADAS and automated vehicle technology or involving claims that involved vehicles should have had ADAS or automated vehicle technology.

Data. Higher levels of automation not only will provide crash investigators with novel and nuanced situations to evaluate but also will provide new forms of data to assist in understanding an incident. For example, traditional accident-reconstruction methodologies will benefit from an enriched data set coming from the various sensor suites and data recorders that feed the ADAS and automated technologies.

With respect to available postincident data, the evolution of event data recorders (EDRs) and data loggers will provide useful information to accident-reconstruction specialists in determining precrash and postcrash maneuvers and movements even in the absence of physical evidence left on the roadway. The Code of Federal Regulations specifies current EDR requirements (if a vehicle is so equipped), including the data elements that have to be recorded before the triggering event (i.e., the crash).23 Currently, vehicles equipped with an EDR are required to record indicated vehicle speed, accelerator pedal position, and service brake application up to five seconds before the triggering event. SAE International’s J1698/1_201805 standard provides a recommended practice, detailing common data output formats and definitions for additional data elements that may be useful for analyzing vehicle crash and crash-like events that meet specified trigger criteria.24 These additional data elements include information on ADAS status and activation that can support the reconstruction of precrash events. Furthermore, SAE International’s EDR committee currently is working on a recommended practice for an ADS (automated driving systems) Data Logger. The draft recommended practice specifies that

[t]he data elements defined in this document are unique to ADS and provide additional background of the events leading up to a collision in the absence of an eye-witness account. The camera(s), LiDAR(s), and other sensor data will provide this eye-witness record.25

The data stored in the ADS Data Logger, in conjunction with the data stored in the EDR, will help accident-reconstruction experts in painting a picture of what happened in the moments leading to a crash (or near-crash), even in the absence of eyewitness and physical evidence.

Human factors. From the perspective of a human factors investigator, the precollision information gleaned from all of the new sources of data in ADAS-equipped vehicles will provide insight into driver behavior that all too often is elusive in current incident investigations. However, for all of the reasons detailed in this article (e.g., consumer understanding, trust in automation, expectation of system performance, false and nuisance alarms, etc.), crashes involving ADAS-equipped vehicles will necessitate an expanded focus on the driver-vehicle interaction. The complex interaction between the driver and level 0–3 systems and the performance of the vehicle for level 0–5 systems will require an expansion and evolution of existing methodologies that examine the role of the perceptual, cognitive, and motor functions of the humans involved in crashes.

In vehicle incidents, a current human factors analysis might include evaluating driver perception-reaction time (PRT)—i.e., the time it takes a driver to perceive an obstacle and produce an appropriate response.26 As ADAS and vehicle automation take over more tasks for the human driver, however, understanding what the driver may need to perceive and react to becomes slightly more complicated. For example, for many level 0–2 technologies, a driver who should still be attending to the roadway may react to the presence of a roadway hazard but may also react to an in-vehicle warning from an ADAS. The investigator must consider carefully how these perceptions and reactions may influence one another.

What cannot be lost to the investigator looking at driver behavior in the presence of ADAS and vehicle automation is how and to what a driver could have responded in the absence of the technology. For example, was the hazard available to be seen by the driver? As mentioned above, the general appearance, size, shape, and type of object may affect the ability of ADAS to detect it due to limitations of the technology’s sensor suite—and the same is true of the human driver. Features such as object size, uniqueness, contrast, conspicuity, and expectation will factor into the likelihood of detection by a human driver27 and can influence PRT.28

These features are the primary characteristics of the visual information available to a sensor, whether that sensor is an eye or a camera. While the human eye has certain advantages over camera sensors (e.g., dynamic range), the human eye can perceive only what is in the observer’s field of view and also can be affected by goal-directed attention.29 Camera sensors, on the other hand, can expand the field of view available for detection and do not necessarily lose acuity in the periphery. Understanding the capabilities and limitations of the various available perceptual systems (i.e., human or technological sensor) will be essential when looking at human factors aspects of incidents involving ADAS and automated vehicle technology.

As noted above, ADAS technology is effective when it behaves in line with how a human would be expected to behave.30 This provides important constraints for individuals who design, implement, and use ADAS. And in terms of vehicle accident investigations, human factors practitioners will play a critical role by evaluating whether the behavior of an automated or driver-assistance system did, in fact, act in line with the expectations of human drivers.

In addition to looking at the general behavior of the driver in investigations of incidents involving ADAS-equipped vehicles, human factors investigators also must consider the aspects of the technology and its interaction with the driver. ADAS technologies rely on core human factors principles but necessarily create novel driver-vehicle interactions. As an example, imagine a scenario where a forward collision warning occurs, followed by possible activation of AEB without driver intervention. First, the driver will need to identify and understand the warning signal and then decide how or if to respond to it. Possible responses include beginning to search for what set off the warning (i.e., look for the hazard) or to brake or steer. Each of these responses will take time. Our research suggests that many drivers are able to use a forward collision warning to react to a conspicuous hazard (i.e., vehicle target) in their path and avoid a collision.31 Other studies have shown that for inattentive or distracted drivers, it can take a minimum of two to three seconds to acknowledge that action needs to be taken in response to a warning.32 There are many other variables, such as driver expectations, that may affect the takeover time. When a driver has lost situational awareness or is “out of the loop,” it can be quite difficult for that individual to respond to sudden emergencies or warnings.33 While the specific interaction between a human and ADAS is nuanced and situation dependent, one aspect that is clear is that the argument about the effectiveness of these technologies is far more complicated than simply their presence or absence.

Conclusion

Although still in small numbers, ADAS and automated vehicle technologies are here, and all predictions indicate that they will increase in proliferation and complexity in the upcoming years. However, it still will be some time before most vehicles on the road will be equipped with some form of an advanced automated technology. Until a point where all (or at least most) vehicles are automated, the safety and mobility benefits of these systems will not be realized fully. With time, though, a reduction in crashes and incidents is expected.

Regardless, with the increasing complexity of vehicle technology and the corresponding driver-vehicle interactions, multidisciplinary teams of experts from fields including, but not limited to, vehicle accident reconstruction, automotive engineering, human factors, biomechanics, and computer science will be needed to investigate incidents involving ADAS and automated vehicles. In performing these investigations, it is inappropriate simply to assume that the presence of a given technology could have avoided (or even significantly mitigated) a collision (or, conversely, that its presence negatively influenced the outcome). For all of the reasons discussed in this article, careful and detailed investigation must be used to understand what technologies reasonably could have been expected to be available (or not available), the capabilities and limitations of the advanced technologies, and how the driver may have interacted with the technologies. The rapidly changing landscape of vehicle technology and automation and consumers’ understanding of the technologies also mean that each incident and each technology must be evaluated and investigated independently.

Notes

1. NHTSA-IIHS Announcement on AEB, NHTSA.gov (Dec. 21, 2017), www.nhtsa.gov/press-releases/nhtsa-iihs-announcement-aeb.

2. Raja Parasuraman, Thomas B. Sheridan & Christopher D. Wickens, A Model for Types and Levels of Human Interaction with Automation, 30(3) IEEE Transactions on Sys., Man, and Cybernetics—Part A: Sys. & Hums. 286, 291 (2000).

3. IIHS, Most Honda Owners Turn Off Lane Departure Warning, 51(1) Status Rep. 1, 6 (2016).

4. Tara A. Kazi, Neville A. Stanton, Guy H. Walker & Mark S. Young, Designer Driving: Drivers’ Conceptual Models and Level of Trust in Adaptive Cruise Control, 45(3) Int’l J. Vehicle Design 339 (2007).

5. Kristin E. Schaefer, Jessie Y. C. Chen, James L. Szalma & P. A. Hancock, A Meta-Analysis of Factors Influencing the Development of Trust in Automation: Implications for Understanding Autonomy in Future Systems, 58(3) Hum. Factors 377 (2016).

6. John L. Campbell, Christian M. Richard, James L. Brown & Marvin McCallum, Nat’l Highway Traffic Safety Admin., DOT HS 810 697, Crash Warning System Interfaces: Human Factors Insights and Lessons Learned 2–10 (2007); John L. Campbell et al., Nat’l Highway Traffic Safety Admin., DOT HS 812 360, Human Factors Design Guidance for Driver-Vehicle Interfaces 4–2 (2016).

7. Genya Abe & John Richardson, The Influence of Alarm Timing on Driver Response to Collision Warning Systems Following System Failure, 25(5) Behav. & Info. Tech. 443, 448 (2006).

8. James R. Sayer et al., Nat’l Highway Traffic Safety Admin., Integrated Vehicle-Based Safety Systems: Light-Vehicle Field Operational Test Key Findings Report 3, 5 (2011).

9. Id.

10. Emily Nodine, Andy Lam, Scott Stevens, Michael Razo & Wassim Najm, Nat’l Highway Traffic Safety Admin., DOT HS 811 516, Integrated Vehicle-Based Safety Systems (IVBSS): Light Vehicle Field Operational Test Independent Evaluation 58 (2011).

11. Monica N. Lees & John D. Lee, The Influence of Distraction and Driving Context on Driver Response to Imperfect Collision Warning Systems, 50 Ergonomics 1264, 1274 (2007).

12. Id.

13. Deborah A. Collard & Nigel L. Mortimer, A Survey of Canadian Drivers’ Knowledge About and Experience with Anti-Lock Brakes, in Proceedings: Int’l Technical Conf. on Enhanced Safety of Vehicles 422–27 (Nat’l Highway Traffic Safety Admin. 1998); Allan F. Williams & JoAnn K. Wells, Driver Experience with Antilock Brake Systems, 26(6) Accident Analysis & Prevention 807–11 (1994).

14. Christian Hoyos, Benjamin D. Lester, Caroline Crump, David M. Cades & Doug Young, Consumer Perceptions, Understanding, and Expectations of Advanced Driver Assistance Systems (ADAS) and Vehicle Automation, in 62(1) Proceedings of the Hum. Factors & Ergonomics Soc’y Ann. Meeting 1888–92 (SAGE Publ’ns Sept. 2018).

15. Id.

16. Michael A. Goodrich, Erwin R. Boer & Hideaki Inoue, A Model of Human Brake Initiation Behavior with Implications for ACC Design, Proceedings 199 IEE/IEEJ/JASI Int’l Conf. on Intelligent Transp. Sys. 86, 90 (1999).

17. Trent W. Victor, Emma Tivesten, Par Gustavsson, Joel Johansson, Frederick Sangberg & Mikael L. Aust, Automation Expectation Mismatch: Incorrect Prediction Despite Eyes on Threat and Hands on Wheel, 60(8) Hum. Factors 1095, 1113 (2018).

18. Michael A. Goodrich & Erwin R. Boer, Model-Based Human-Centered Task Automation: A Case Study in ACC System Design, 33(3) IEEE Transactions on Sys., Man, & Cybernetics—Part A: Sys. & Hums. 325, 328 (2003).

19. Goodrich et al., supra note 16.

20. Richard E. Redding, John R. Cannon & Thomas L. Seamster, Expertise in Air Traffic Control (ATC): What Is It, and How Can We Train for It?, 36 Proceedings Hum. Factors & Ergonomics Soc’y 1326, 1327 (1992).

21. Andrea M. Cassidy, Mental Models, Trust, and Reliance: Exploring the Effect of Human Perceptions on Automation Use (2009) (master’s thesis, Naval Postgraduate School, Monterey, California); John D. Lee & Katrina A. See, Trust in Automation: Designing for Appropriate Reliance, 46(1) Hum. Factors 50, 55 (2004).

22. Bart D. Wilkison, Effects of Mental Model Quality on Collaborative System Performance 45 (2008) (master’s thesis, Georgia Institute of Technology).

23. Event Data Recorders, 49 C.F.R. § 563 (2011).

24. SAE Int’l, SAE Standard J1698, Event Data Recorder: Output Data Definition (rev. May 2018).

25. SAE Int’l, SAE Standard J3197 (WIP), ADS Data Logger 1 (Jan. 2012).

26. David Krauss, Forensic Aspects of Driver Perception and Response 233, 236–37 (Lawyers & Judges Publishing Co. 4th ed. 2015).

27. For a review of these and related topics, see id. at 57, 58.

28. Martin P. Langham & Nicholas J. Moberly, Pedestrian Conspicuity Research: A Review, 46(4) Ergonomics 345, 358 (2003); Thomas J. Triggs & Walter G. Harris, Reaction Time of Drivers to Road Stimuli 1, 45 (June 1982) (Human Factors Report No. HFR-12).

29. Daniel J. Simons & Christopher F. Chabris, Gorillas in Our Midst: Sustained Inattentional Blindness for Dynamic Events, 28(9) Perception 1059 (1999).

30. Goodrich et al., supra note 16.

31. Caroline Crump, David Cades, Robert Rauschenberger, Emily Hildebrand, Jeremy Schwark, Brandon Barakat & Douglas Young, Driver Reactions in a Vehicle with Collision Warning and Mitigation Technology, 59(1) Proceedings Hum. Factors & Ergonomics Soc’y Ann. Meeting 1321, 1325 (2015).

32. Christian Gold, Daniel Damböck, Lutz Lorenz & Klaus Bengler, “Take Over!” How Long Does It Take to Get the Driver Back into the Loop?, 57(1) Proceedings Hum. Factors & Ergonomics Soc’y 57th Ann. Meeting 1938, 1940 (2013); David Miller, Annabel Sun, Mishel Johns, Hillary Ive, David Sirkin, Sudipto Aich & Wendy Ju, Distraction Becomes Engagement in Automated Driving, 59(1) Proceedings Hum. Factors & Ergonomics Soc’y Ann. Meeting 1676, 1678 (Sept. 2015); Jonas Radlmayr, Christian Gold, Lutz Lorenz, Mehdi Farid & Klaus Bengler, How Traffic Situations and Non-Driving Related Tasks Affect the Take-Over Quality in Highly Automated Driving, 58(1) Proceedings Hum. Factors & Ergonomics Soc’y Ann. Meeting 2063, 2065–66 (Sept. 2014).

33. Mica R. Endsley & Esin O. Kiris, The Out-of-the-Loop Performance Problem and Level of Control in Automation, 37(2) Hum. Factors: J. Hum. Factors & Ergonomics Soc’y 381, 391 (1995).

Download a PDF of figures for this article.

Entity:
Topic:

David Cades is a principal scientist in the Human Factors Practice at Exponent in Chicago, Illinois. He specializes in human factors, driver behavior, automated vehicles, and advanced driver-assistance systems (ADAS).

Carmine Senatore is a manager in the Vehicle Engineering Practice at Exponent in Natick, Massachusetts. His specialties include accident reconstruction, functional safety, automated vehicles, and ADAS.

John L. Campbell is a senior managing scientist in the Human Factors Practice at Exponent in Bellevue, Washington. His areas of expertise include human factors, roadway safety and design, connected and automated vehicles, and advanced driver-vehicle interfaces.

Ryan Harrington is a principal in the Vehicle Engineering Practice at Exponent in Natick, Massachusetts. He specializes in development and deployment of emerging automotive technologies, ADAS, and automated vehicles.

Daniel Wood is a scientist in the Human Factors Practice at Exponent in Chicago, Illinois. His specialties include human factors, driver behavior, and visual perception and attention.