Quantifying drivers’ perceived risk is important in the design and evaluation of the behaviour of automated vehicles (AVs) and in predicting takeovers by the driver. A ‘Driver's Risk Field’ (DRF) function has been previously shown to be able to predict manual driving behaviour in several simulated scenarios. In this paper, we tested if the DRF-based risk estimate (rˆ) could predict manual driving behaviour and the driver's perceived risk during automated driving. To ensure that the participants perceived realistic levels of risk, the experiment was conducted in a test vehicle. Eight participants drove five laps manually and experienced 12 different laps of automated driving on a test track. The test track consisted of three sections (which were sub-divided into 12 sectors): curve driving (9 sectors), parked car (1 sector), and 90-degree intersections (2 sectors). If the driver verbally expressed risk or performed a takeover, that particular sector was labelled as risky. The results show that the DRF risk estimate (rˆ) predicted manual driving behaviour (ρ_steering=0.69, ρ_speed=0.64), as well as correlated with the driver's perceived risk in curve driving (r² = 0.98) and while negotiating a car parked outside the lane boundary (r²=0.59). In conclusion, the DRF-based risk estimate (rˆ) is predictive of manual driving behaviour and perceived risk in automated driving. Future research should include tactical and strategic components to the driving task.

Did you know that most drivers swing left before taking a right curve? In fact, this is a given for all race car drivers and a rule for efficient curve negotiation. This distinct way of approaching a curve is called prepositioning. In a recent study it is found that incorporating knowledge of this prepositioning phase is crucial for the reliability and acceptance of some trajectory-guiding advanced-driver-assistance-systems. Unfortunately, our understanding of prepositioning behaviour is still limited, with most driver models unable to account for this phenomenon. In an attempt to improve our understanding, the effects of changing velocity and road radius on prepositioning behaviour are studied experimentally in a fixed-base driving simulator. Twenty-four participants drove four conditions comprising two different fixed-speed velocities (50 and 80 km/h) and two different curve radii (204 and 350 m). The results show that the drivers' maximum prepositioning position significantly increases with increasing velocity and significantly decreases with increasing radius. With 88% of the runs exhibiting a significant displacement, i.e. larger than 0.05 m relative to a constant road bias. The findings suggest that drivers adjust their prepositioning behaviour to the Time-to-Line-Crossing (TLC) of the road environment, in an attempt to maximise TLC. Incorporating these findings in future driver modelling will bridge the gap between straight road and in-curve driving behaviour, thereby bolstering the descriptive capacity of these models.

When taking a curve, drivers follow their own unique trajectory. Most driver style classifiers in literature are based on inertial inputs, denoting whether a given driver is aggressive or calm. However, this does not give any indication of a drivers trajectory style, i.e. whether a driver is curve cutting. To fill this void, this paper introduces a novel rule based classifier that categorises seven different trajectory styles. The classifier is applied to data from a fixed-base driving simulator study in which 45 subjects drove on three roads, comprising three different velocities: 25, 50 and 80 km/h, with three corresponding radii: 20, 80 and 204 m. The results show that some classes are more prevalent than others, with biased outer curve negotiation performed by a majority of the subjects and with no drivers classified as centerline drivers. The proposed trajectory classifier is shown to exhibit high levels of consistency, with 93% of drivers exhibiting consistent trajectory classes for at least 66% of the right curves driven and 84% exhibits consistent trajectory classes for atleast 66% of the left curves driven. Where this consistency indicates a potential for generalising the classification results to other curves. Additionally, this classifier can be used to adapt trajectory-driven advanced driver assistance systems, thereby serving as an alternative to driver modelling.

We conceptually replicated three highly cited experiments on speed adaptation, by measuring drivers’ experienced risk (galvanic skin response; GSR), experienced task difficulty (self-reported task effort; SRTE), and safety margins (time-to-line-crossing; TLC) in a single experiment. The three measures were compared using a nonparametric index that captures the criteria of constancy during self-paced driving and sensitivity during forced-paced driving. In a driving simulator, 24 participants completed two forced-paced and one self-paced run. Each run held four different lane width conditions. Results showed that participants drove faster on wider lanes, thus confirming the expected speed adaptation. None of the three measures offered persuasive evidence for speed adaptation because they failed either the sensitivity criterion (GSR) or the constancy criterion (TLC, SRTE). An additional measure, steering reversal rate, outperformed the other three measures regarding sensitivity and constancy, prompting a further evaluation of the role of control activity in speed adaptation.

Shared control is an increasingly popular approach to facilitate control and communication between humans and intelligent machines. However, there is little consensus in guidelines for design and evaluation of shared control, or even in a definition of what constitutes shared control. This lack of consensus complicates cross fertilization of shared control research between different application domains. This paper provides a definition for shared control in context with previous definitions, and a set of general axioms for design and evaluation of shared control solutions. The utility of the definition and axioms are demonstrated by applying them to four application domains: automotive, robot-assisted surgery, brain–machine interfaces, and learning. Literature is discussed for each of these four domains in light of the proposed definition and axioms. Finally, to facilitate design choices for other applications, we propose a hierarchical framework for shared control that links the shared control literature with traded control, co-operative control, and other human–automation interaction methods. Future work should reveal the generalizability and utility of the proposed shared control framework in designing useful, safe, and comfortable interaction between humans and intelligent machines.

A body illusion, commonly known in the form of the “Rubber Hand Illusion”, is an illusion wherein visual inputs on an inanimate object and simultaneous tactile inputs on a part of the body lead to a situation where the inanimate object is identified as the body part. This study investigated the possibility of inducing a body illusion during a teleoperated reaching task, to see if this leads to increased telepresence and improved accuracy. Three conditions were presented in random order; the Direct Control (DC) condition, where the participant’s hand is shown on the screen, the Projected Hand Illusion (PHI) condition, showing the slave device consisting of a 3D-printed hand designed to induce a body illusion, and the no Projected Hand Illusion (nPHI) condition, showing the slave device consisting of a 3D-printed object of appropriate shape but designed to not induce a body illusion. Reaching performance was interpreted in terms of position error, for which a significant difference was found between conditions PHI and nPHI. In the nPHI condition, participants kept more distance to the obstacle than in the PHI condition. Potential causes for this difference are an increased perception of risk due to a difference in visual perception, or subtle visual differences in between the two conditions.

Manual control cybernetics aims to understand and describe how humans control vehicles and devices using mathematical models of human control dynamics. This “cybernetic approach” enables objective and quantitative comparisons of human behavior, and allows a systematic optimization of human control interfaces and training associated with manual control. Current cybernetics theory is primarily based on technology and analysis methods formalized in the 1960s and has shown to be limited in its capability to capture the full breadth of human cognition and control. This paper reviews the current state-of-the-art in our knowledge of human manual control, points out the main fundamental limitations in cybernetics, and proposes a possible roadmap to advance the theory and its applications. Central in this roadmap will be a shift from the current linear time-invariant modeling approach that is only truly valid for human behavior under tightly controlled and stationary conditions, to methods that facilitate the analysis of adaptive, and possibly time-varying, human behavior in realistic control tasks. Examples of key current developments in the field of cybernetics—human use of preview, predictable discrete maneuvering, skill acquisition and training, time-varying human modeling, and neuromuscular system modeling—that contribute to this shift are presented in this paper. The new foundations for cybernetics that will emerge from these efforts will impact all domains that involve humans in manual and semiautomatic control.

Manual control cybernetics aims to understand and describe how humans control vehicles and devices, such that more effective human-machine interfaces can be designed. Current cybernetics theory is primarily based on technology and analysis methods developed in the 1960s and has shown to be limited in its capability to capture the full breadth of human cognition and control. This paper summarizes some of the main fundamental limitations in cybernetics and provides a possible road-map to advance the theory and its applications. Central in this agenda will be a shift from the current linear time-invariant modeling approach, to the use of linear parameter-varying system models. Recent progress in identification methods of these latter models may allow us, for the first time, to mathematically model and identify time-varying, adaptive human control, opening up many opportunities to systematically optimize our human control interfaces and training. New foundations for cybernetics will impact all domains that involve humans in manual and semi-automatic control.

Driver training effectiveness requires assessment of driver performance in order to compare and contrast the impact of different training techniques on the learned control. Human performance can be quantified from different perspectives ranging from aggregate measures to specific model coefficients in order to link observed performance to how the driver achieved this performance through a particular control strategy. A model based approach is needed to understand the pros and cons of different training programs. One important yet often ignored aspect of a model is the cost function that drives behavior adaptation. Here, a model based methodology is proposed that estimates the weights on different terms in the cost function that drivers use to adapt their behavior in order to satisfy their performance needs. Because the driver model includes the effect of the controlled dynamical system as well as any particularities of the training environment, one can use it to quantify the effect of training specific deviations from reality, such as the use of a driving simulator that causes known biases in perception, on behavior. This paper details the methodological approach and discusses it in the context of stopping behavior in reality versus in a driving simulator. The goal with training is to instill the right structural behavior so that only minor adaptations may be needed once applying the learned skill in reality. Because the adopted cost function plays such a large role, much focus should also be given to shaping the cost function that operators employ.

Drivers exhibit a range of lateral positions that depend on their location down the road in relationship to the curves. For example, within a curve the range is narrower and biased towards the inner road edge compared to straight sections where it is more centralized and wider. The question is what shapes the swath of lateral positions that drivers accept. It is generally accepted that drivers under normal conditions (i.e. not race car drivers) are satisficing controllers rather than optimizing controllers that operate based on a number of aspiration levels that they try to protect. In this paper, lower bounds on straight and curved time to line crossing (TLC) are hypothesized as aspiration levels that determine the acceptable range of lateral offsets. These lower TLC bounds also shape the speed at which a curve can be taken because when the aspiration levels cannot be reached at the current speed, a slower speed needs to be adopted to maintain sufficiently high straight and/or curved TLCs. This paper provides a theoretical exposition of straight and curved TLC as a function of environmental and driver skill factors. It provides much evidence that TLC plays a role in mediating drivers’ lateral position and speed choices.

Driving is a tracking task with preview as has been recognized since the 60s. Subsequent research to model human curve negotiation divides into two camps. One in which a limited number of points in the future (generally one or two) are used to guide lane keeping control on straight and curved roads and another that uses optimal preview control (OPC) to characterize human control behavior. The former is too simplistic as it cannot accurately handle curve entry and exit with a single preview and gain setting (i.e. non situation adaptive) and the latter is arguably too computationally intense for a human to adopt (but not unreasonable to converge to over time). This paper shows theoretically that by selecting two preview points strategically related to vehicle dynamics for near preview point and striking a balance between curve cutting on entry, curve overshoot on exit, and smooth control throughout for the far preview point, two-point-controllers approach the performance of a full optimal preview controller. The difference between the reference full OPC and two-point-controllers lies mainly in the fact that the three phases of curve negotiation (entry, within, and exit) require different previews and gains which only the OPC is capable of.

Purpose: To evaluate a new test paradigm for saccadic and smooth pursuit eye movement performance during an attentional and visual performance-based task.

Purpose: To examine the relationship between Motor Vehicle Collisions (MVCs) in
drivers with glaucoma and standard automated perimetry (SAP), Useful Field of View
(UFOV), and driving simulator assessment of divided attention.
Methods: A cross-sectional study of 153 drivers from the Diagnostic Innovations in
Glaucoma Study. All subjects had SAP and divided attention was assessed using UFOV
and driving simulation using low-, medium-, and high-contrast peripheral stimuli
presented during curve negotiation and car following tasks. Self-reported history of
MVCs and average mileage driven were recorded.
Results: Eighteen of 153 subjects (11.8%) reported a MVC. There was no difference in
visual acuity but the MVC group was older, drove fewer miles, and had worse
binocular SAP sensitivity, contrast sensitivity, and ability to divide attention (UFOV and
driving simulation). Low contrast driving simulator tasks were the best discriminators
of MVC (AUC 0.80 for curve negotiation versus 0.69 for binocular SAP and 0.59 for
UFOV). Adjusting for confounding factors, longer reaction times to driving simulator
divided attention tasks provided additional value compared with SAP and UFOV, with
a 1 standard deviation (SD) increase in reaction time (approximately 0.75 s) associated
with almost two-fold increased odds of MVC.
Conclusions: Reaction times to low contrast divided attention tasks during driving
simulation were significantly associated with history of MVC, performing better than
conventional perimetric tests and UFOV.
Translational Relevance: The association between conventional tests of visual
function and MVCs in drivers with glaucoma is weak, however, tests of divided
attention, particularly using driving simulation, may improve risk assessment.

Steer-by-Wire (SbW) systems currently under development by the automotive industry offer interesting new approaches to designing driver-steering wheel interactions. The traditional, emerging dynamics in mechanically linked steering systems can be re-designed with SbW to improve or even extend the steering 'feel'. In this article we manipulated the steering wheel dynamics such that each design was expected to yield the best driving performance with the least amount of driver control effort for a particular driving task. We tested three designs during three different driving tasks in a fixed-base driving simulator. The results of the experiment showed that steering wheel dynamics should be stiff and sluggish for driving on straight roads and slack and light for curve negotiation. Future experiments will investigate the implications for drivers on a neuromuscular level.

Negotiating intersections is a complex driving task that is particularly difficult for older drivers. This task requires accurate coordination of multiple driving subtasks, placing high demands on perception, attention and motor control that are known to decline with age. We analyzed intersection negotiation behavior in an instrumented vehicle and found striking differences in how drivers of different ages synchronize speed and heading control when turning right. The older drivers performed most of their steering while standing still instead of while accelerating as younger drivers do. This shift from parallel to serial control is a compensatory solution that drivers employ in response to age related decline in perception, cognition, and motor control abilities. Serialization of turning at an intersection reduces attentional demands largely by eliminating the need to switch attention between different driving sub-tasks.