In navigation applications, Artificial Intelligence (AI) can improve efficiency and decision making. It is not clear, however, how designers should account for human cooperation when integrating AI systems in navigation work. In a novel empirical study, we examine the transition in the maritime domain towards higher levels of machine autonomy. Our method involved interviewing technology designers (n = 9) and navigators aboard two partially automated ferries (n = 5), as well as collecting field observations aboard one of the ferries. The results indicated a discrepancy between how designers construed human-AI collaboration compared to navigators' own accounts in the field. Navigators reflected upon their role as one of 'backup,' defined by ad-hoc control takeovers from the automation. Designers positioned navigators 'in the loop' of a larger control system but discounted the role of in-situ skills and heuristic decision making in all but the most controlled takeover actions. The discrepancy shed light on how integration of AI systems may be better aligned to human cooperation in navigation. This included designing AI systems that render computational activities more visible and that incorporate social cues that articulate human work in its natural setting. Positioned within the field of AI alignment research, the main contribution is a formulation of human-AI interaction design insights for future navigation and control room work.

This paper presents implementation and evaluation of an intraoperative arthroscopic tracker system for research and educational use. The system enables automatic pose and force data exchange between a physical asset and a digital model, which is required in arthroscopic digital twins. Surgical instrument motion tracking is captured using an inertial measurement unit in combination with stereo vision (ZED mini, Stereolabs, USA), and force registration is achieved using a 6 degree-of-freedom force-torque sensor (Nano 25, ATI Industrial Automation, USA). An integration layer in the software system continuously fetches data from the sensors, and transmits data over a network connection. The software visualization layer displays the instrument pose and force readings relative to a 3D anatomical model in real-time (24 Hz). Position accuracy of the system was compared to a Kuka LBR Med 14 R820 collaborative robot, and the total root mean square error was found to be 12.60 mm. A post-operative instrument trajectory map from sampled pose and force/torque data was implemented in Matlab, and demonstrated in an experiment. The system is feasible for use in many research and educational applications.

Continuous measurement of cardiovascular parameters is important for monitoring cardiovascular health. Ballistocardiography is a noninvasive method of recording cardiovascular events. Here, we present a sensor system prototype for recording of the full-body ballistocardiogram in a wearable. An array of soft bladders in each sole are filled with water and connected to barometric pressure sensors. We demonstrate the use of the prototype to estimate the pulse transit time against continuous blood pressure in a validation experiment (n=14). Participants wore the sensor shoes while standing on a reference weight-scale. Simultaneous recordings were taken of the sole pressure arrays, finger-clip photoplethysmography, and continuous blood pressure via the volume-clamp method. Measurements were taken at rest, during cold-pressor intervention for 60 seconds, and 3 minutes following end of intervention. The waveform of the ballistocardiograms captured by the proposed sensor system corresponded well to the simultaneously collected waveforms from the reference weigh-scale. Pulse-transit time estimated from shoe BCG and PPG show inverse correlation to vasoconstriction-induced blood pressure increase. By demonstrating the use of the system to compute a vascular transit time, we show the potential of ballistocardiographic insoles as a wearable sensor interface for cardiovascular monitoring.

Teleoperation in an environment with latency is difficult and highly stressful for human operators, resulting in high cognitive workload and decreased human performance. This work investigates if a simple predictive display can increase performance and lower subjective workload for the human operator when teleoperating a remotely operated vehicle (ROV). A predictive display based on image transformation was developed by applying positional and scale transformations to the video feed and tested. An experiment was designed, consisting of a simple navigational task (peg-in-hole game) with a ground ROV, in three distinct conditions: C1. Latency, C2. Latency with predictive display (PD) and C3. Baseline (no added latency). Findings from N = 57 participants show a statistically significant increase of 20% in human performance with the aid of the predictive display. Although differences in subjective workload was not statistically significant, both subjective performance and actual game performance did increase significantly by using the predictive display. In fact, the latter almost doubled for participants defining themselves as regular gamers. Lastly, A principle component analysis (PCA) was conducted investigating confounding factors with confirmatory results.

Many producers of automated vehicle systems have begun testing autonomous vehicles on the road. In order to ensure safety and prevent crashes, human drivers are enlisted to monitor autonomous vehicles. However, operators of autonomous systems exhibit negative behavior adaptations in response to prolonged supervision of automation. To prevent the onset of undesirable behaviors in safety drivers, we must investigate driver state and behavior changes during the operation of highly automated vehicles. In the study presented here, we examine the effects of theoretical and practical training on the drivers' response to potentially critical situations in a longitudinal driving simulator study. We also present the effects of encountering a failure of the automated vehicle on driver state and behavior. We conducted a two-part panel driving simulator study (N=28), with an interval of 20-30 days between the training and testing sessions. We found that while participants with training are better prepared for a potential failure of the automation, participants in both conditions show a rise in sleepy or drowsy behavior before a potential failure of automation.

The advent of automated features in modern vehicles requires human factors researchers to find measures other than driving behavior to anticipate the response of drivers in various contexts. Functional near-infrared spectroscopy (fNIRS) is one research tool that allows us to quantify the driver's mental state. However, the underlying mechanisms of fNIRS technology can limit the possible contexts for its application. The pervasive question arises, whether the measurement device at hand is suitable for the research topic in question and is it capable of detecting the phenomenon under investigation? We provide a proof of concept study demonstrating that significant habituation is present when drivers operate new automated driving systems and that fNIRS technology is suitable to detect said driver habituation effects. The study presented here was conducted in a driving simulator and investigated the drivers' cortical activation in three different modes of automation: manual, partially autonomous, and fully autonomous modes.

Information about drivers' mental states can be vital to the design of interfaces for highly automated vehicles. Functional near infrared spectroscopy (fNIRS) is a neuroimaging tool that is fast becoming popular to study the cortical activity of participants in HCI experiments and driving simulator studies in particular. The analysis methods of the fNIRS data create requirements in the experimental design such as repeated measures. In this paper, we present a study of the event related cortical activity of the drivers of manual, partially autonomous, and fully autonomous cars when performing lane changes using functional near infrared spectroscopic measures. We also present the experimental methodology that was adopted to meet the needs of the fNIRS measurement and the subsequent analysis. The study (N=28) was conducted in a driving simulator. Participants drove for approximately 7 minutes and performed 8 lane change maneuvers in each mode of automation. Multiple streams of data including 4 time-synced video recordings, NASA TLX questionnaires and fNIRS data were recorded and analyzed. It was found that the dorsolateral prefrontal cortex activation during lane changes performed in a partially autonomous mode of operation was just as high as that during a manual lane change, showing that drivers of partially automated systems are as cognitively engaged as drivers of manually operated vehicles.

We explore the impact of the displayed precision of instrumentation estimates of range and state-of-charge on drivers' attitudes towards an all-electric vehicle (EV), on their driving experience, and driving behavior under varying conditions of resource availability. Participants (N=73) completed a 19-mile long drive through highway, rural town and mountain road conditions in an EV that displayed high vs. low remaining range, and gave estimates of that range with high and low information ambiguity. We found that an ambiguous display of range preserved drivers' feelings of trust towards the vehicle, despite encountering situations intended to induce severe range anxiety. Furthermore, compared to drivers facing an unambiguous display of range, drivers presented with an ambiguous range display reported improved driving experience, and exhibited driving behavior better adapted to road and remaining range conditions.