Erwin R. Boer

Haptic shared control: smoothly shifting control authority?

Literature points to persistent issues in human-automation interaction, which are caused either when the human does not understand the automation or when the automation does not understand the human. Design guidelines for human-automation interaction aim to avoid such issues and commonly agree that the human should have continuous interaction and communication with the automation system and its authority level and should retain final authority. This paper argues that haptic shared control is a promising approach to meet the commonly voiced design guidelines for human-automation interaction, especially for automotive applications. The goal of the paper is to provide evidence for this statement, by discussing several realizations of haptic shared control found in literature. We show that literature provides ample experimental evidence that haptic shared control can lead to short-term performance benefits (e.g., faster and more accurate vehicle control; lower levels of control effort; reduced demand for visual attention). We conclude that although the continuous intuitive physical interaction inherent in haptic shared control is expected to reduce long-term issues with human-automation interaction, little experimental evidence for this is provided. Therefore, future research on haptic shared control should focus more on issues related to long-term use such as trust, overreliance, dependency on the system, and retention of skills.

Designing human-centered automation: trade-offs in collision avoidance system design

Human-centered automation problems have multiple attributes: an attribute reflecting human goals and capabilities, and an attribute reflecting automation goals and capabilities. In the absence of a general theory of human interaction with complex systems, it is difficult to define and find a unique optimal multiattribute resolution to these competing design requirements. We develop a systematic approach to such problems using a multiattribute decomposition of human and automation goals. This paradigm uses both the satisficing decision principle which is unique to two-attribute problems, and the domination principle which is a common manifestation of the optimality principle in multiattribute domains. As applied to human-centered automation in advanced vehicle systems, the decision method identifies performance evaluations and compares the safety benefit of a system intervention against the cost to the human operator. We illustrate the method by analyzing an automated system to prevent lane departures.

The effect of haptic guidance on curve negotiation behavior of young, experienced drivers

Haptic feedback on the steering wheel is reported in literature as a promising way to support drivers during steering tasks. Haptic support allows drivers to remain in the direct manual control loop, avoiding known human factors issues with automation. This paper proposes haptic guidance based on the concept of shared control, where both the driver and the support system influence the steering wheel torque. The haptic guidance is developed to continuously generate relatively low forces on the steering wheel, requiring the drivers active steering input to safely negotiate curves. An experiment in a fixed-base driving simulator was conducted, in which 12 young, experienced drivers steered a vehicle - with and without haptic guidance - at a fixed speed along a road with varying curvature. The haptic guidance allowed drivers to slightly but significantly improve safety boundaries in their curve negotiation behavior. Their steering activity was reduced and smoother. The results indicated that continuous haptic guidance is a promising way to support drivers in actively producing (more) optimal steering actions during curve negotiation.

Sharing Control With Haptics: Seamless Driver Support From Manual to Automatic Control

Objective: Haptic shared control was investigated as a human–machine interface that can intuitively share control between drivers and an automatic controller for curve negotiation. Background: As long as automation systems are not fully reliable, a role remains for the driver to be vigilant to the system and the environment to catch any automation errors. The conventional binary switches between supervisory and manual control has many known issues, and haptic shared control is a promising alternative. Method: A total of 42 respondents of varying age and driving experience participated in a driving experiment in a fixed-base simulator, in which curve negotiation behavior during shared control was compared to during manual control, as well as to three haptic tunings of an automatic controller without driver intervention. Results: Under the experimental conditions studied, the main beneficial effect of haptic shared control compared to manual control was that less control activity (16% in steering wheel reversal rate, 15% in standard deviation of steering wheel angle) was needed for realizing an improved safety performance (e.g., 11% in peak lateral error). Full automation removed the need for any human control activity and improved safety performance (e.g., 35% in peak lateral error) but put the human in a supervisory position. Conclusion: Haptic shared control kept the driver in the loop, with enhanced performance at reduced control activity, mitigating the known issues that plague full automation. Application: Haptic support for vehicular control ultimately seeks to intuitively combine human intelligence and creativity with the benefits of automation systems.

Model-based human-centered task automation: a case study in ACC system design

Engineers, business managers, and governments are increasingly aware of the importance and difficulty of integrating technology and humans. The presence of technology can enhance human comfort, efficiency, and safety, but the absence of human-factors analysis can lead to uncomfortable, inefficient, and unsafe systems. Systematic human-centered design requires a basic understanding of how humans generate and manage tasks. A very useful model of human behavior generation can be obtained by recognizing the task-specific role of mental models in not only guiding execution of skills but also managing initiation and termination of these skills. By identifying the human operators mental models and using them as templates for automating different tasks, we experimentally support the hypothesis that natural and safe interaction between human operator and automation is facilitated by this model-based human-centered approach. The design of adaptive cruise control (ACC) systems is used as a case study in the design of model-based task automation systems. Such designs include identifying ecologically appropriate perceptual states, identifying perceptual triggering events for managing transitions between skilled behaviors, and coordinating the actions of automation and operator.

Measuring Neuromuscular Control Dynamics During Car Following With Continuous Haptic Feedback

In previous research, a driver support system that uses continuous haptic feedback on the gas pedal to inform drivers of the separation to the lead vehicle was developed. Although haptic feedback has been previously shown to be beneficial, the influence of the underlying biomechanical properties of the driver on the effectiveness of haptic feedback is largely unknown. The goal of this paper is to experimentally determine the biomechanical properties of the ankle-foot complex (i.e., the admittance) while performing a car-following task, thereby separating driver responses to visual feedback from those to designed haptic feedback. An experiment was conducted in a simplified fixed-base driving simulator, where ten participants were instructed to follow a lead vehicle, with and without the support of haptic feedback. During the experiment, the lead vehicle velocity was perturbed, and small stochastic torque perturbations were applied to the pedal. Both perturbations were separated in the frequency domain to allow the simultaneous estimation of frequency response functions of both the car-following control behavior and the biomechanical admittance. For comparison to previous experiments, the admittance was also estimated during three classical motion control tasks (resist forces, relax, and give way to forces). The main experimental hypotheses were that, first, the haptic feedback would encourage drivers to adopt a “give way to force task,” resulting in larger admittance compared with other tasks and, second, drivers needed less control effort to realize the same car-following performance. Time- and frequency-domain analyses provided evidence for both hypotheses. The developed methodology allows quantification of the range of admittances that a limb can adopt during vehicle control or while performing a variety of motion control tasks. It thereby allows detailed computational driver modeling and provides valuable information on how to design and evaluate continuous haptic feedback systems.

Motivation for continuous haptic gas pedal feedback to support car following

The last years, increased effort has been dedicated to the design of systems that assist the driver in car following. The need for assistance systems arises from the fallibility of the visual feedback loop, for example due to inattention. Existing driver assistance systems either automate the car-following task or support drivers with binary warning systems to redirect their attention when necessary. The goal of this paper is to discuss the benefits and limitations of these systems, and to show the possibilities of an alternative design approach. To attain the goal, a theoretic analysis is presented, that views car following as a closed-loop control task that requires sufficient feedback about the separation (relative distance, relative velocity) to a lead vehicle. A task analysis helps to identify the areas where the current systems assist the driver well, and where they do not. The new design approach aims to keep the human in the loop, by supplementing the semi-continuous visual feedback loop with an additional continuous feedback loop, namely haptic feedback applied directly at the gas pedal. Expected benefits compared to existing systems include: better situation awareness (even during periods of visual inattention) and faster responses (the haptic feedback is available directly at the gas pedal, allowing the use of fast reflexes). Several design issues are presented, such as the prevention of nuisance and fatigue, deciding which separation states the feedback is based upon, and challenges in determining the correct characteristics of the haptic signals. The benefits of the approach are illustrated through several examples from literature that describe experimental humanin-the-loop studies with continuous haptic feedback. It is concluded that haptic feedback on the gas pedal is a promising way of supporting drivers.

Association Between Progressive Retinal Nerve Fiber Layer Loss and Longitudinal Change in Quality of Life in Glaucoma

IMPORTANCE Evaluation of structural optic nerve damage is a fundamental part of diagnosis and management of glaucoma. However, the relationship between structural measurements and disability associated with the disease is not well characterized. Quantification of this relationship may help validate structural measurements as markers directly relevant to quality of life. OBJECTIVE To evaluate the relationship between rates of retinal nerve fiber layer (RNFL) loss and longitudinal changes in quality of life in glaucoma. DESIGN, SETTING, AND PARTICIPANTS Observational cohort study including 260 eyes of 130 patients with glaucoma followed up for a mean (SD) of 3.5 (0.7) years. All patients had repeatable visual field defects on standard automated perimetry (SAP) at baseline. The 25-item National Eye Institute Visual Function Questionnaire (NEI VFQ-25) was performed annually, and spectral-domain optical coherence tomography and SAP were performed at 6-month intervals. A joint model was used to investigate the association between change in NEI VFQ-25 Rasch-calibrated scores and change in RNFL thickness, adjusting for confounding socioeconomic and clinical variables. MAIN OUTCOMES AND MEASURES Association between change in binocular RNFL thickness (RNFL thickness in the better eye at each point) and change in NEI VFQ-25 scores. RESULTS Progressive binocular RNFL thickness loss was associated with worsening of NEI VFQ-25 scores over time. In a multivariable model adjusting for baseline disease severity and the rate of change in binocular SAP sensitivity, each 1-μm-per-year loss of RNFL thickness was associated with a decrease of 1.3 units (95% CI, 1.02-1.56) per year in NEI VFQ-25 scores (P < .001). After adjusting for the contribution from SAP, 26% (95% CI, 12%-39%) of the variability of change in NEI VFQ-25 scores was associated uniquely with change in binocular RNFL thickness. The P value remained less than .001 after adjusting for potential confounding factors. CONCLUSIONS AND RELEVANCE Progressive binocular RNFL thickness loss was associated with longitudinal loss in quality of life, even after adjustment for progressive visual field loss. These findings suggest that rates of binocular RNFL change are valid markers for the degree of neural loss in glaucoma with significant relationship to glaucoma-associated disability.

Effects of alcohol impairment on motorcycle riding skills

Alcohol intoxication is a significant risk factor for fatal traffic crashes; however, there is sparse research on the impairing effects of alcohol on skills involved in motorcycle control. Twenty-four male motorcycle riders between the ages of 21 and 50 were assessed on a test track with task scenarios based on the Motorcycle Safety Foundations (MSF) training program. A balanced incomplete block design was used to remove confounding artifacts (learning effects) by randomizing four BAC levels across three test days. In general, intoxicated riders demonstrated longer response times and adopted larger tolerances leading to more task performance errors. Most of the alcohol effects were evident at the per se 0.08% alcohol level, but some of the effects were observed at the lower 0.05% alcohol level. The effects of alcohol on motorcycle control and rider behavior were modest and occurred when task demand was high (offset weave), time pressure was high (hazard avoidance for near obstacles), and tolerances were constrained (circuit track). The modest effects may be due to the study design, in which experienced riders performed highly practiced, low-speed tasks; alcohol at these levels may produce larger effects with less experienced riders in more challenging situations.

Driving Simulation as a Performance-based Test of Visual Impairment in Glaucoma

The fundamental goal of glaucoma management is to prevent patients from developing visual impairments sufficient to produce disability in their daily lives and impair their quality of life. Ultimately, patients are interested in how their vision will impact their ability to perform daily activities, such as driving. Although technologic advancements such as automated perimetry and devices for optic nerve imaging have resulted in great improvement in our ability to quantify structural and functional damage in glaucoma, the impact on vision-related quality of life of some of the information acquired from these tests remain elusive. In contrast, performance-based measures may be better correlated to traditional measures of vision health and, more importantly, they provide a more direct measure of disability. Driving simulators can be used as a performance-based test for evaluation of functional impairment in glaucoma. Their use can potentially help in the evaluation of driving safety and performance of diseased subjects and provide insight into the different mechanisms involved in causing driving impairment in this disease. The ability to do this in an experimentally controlled and standardized setting enables testing of a much larger number of hypotheses compared with on-road evaluations. Besides evaluating driver fitness, simulators could also potentially be used as a sophisticated test to evaluate cognitive impairment in the context of an everyday task (driving) that has not been available through traditional neuropsychologic assessment.