Search for: All records

Abstract This paper presents the development of an interface that enables humans to collaborate with robot swarms (RSs) to effectively accomplish missions in highly uncertain environments. Autonomous RS technology has advanced to the point where robots can perform various tasks without human intervention. However, there are still two major uncertainties in mission execution. The first uncertainty arises from the limitations of artificial intelligence (AI), such as the inaccuracy of computer vision systems. The second uncertainty comes from the decision-making aspect, where the mission requirements may not be accurately communicated to the RS. To overcome these uncertainties, it is crucial to design a system that allows humans to monitor the RS and actively participate in critical decision-making processes during missions. We design an interface to: 1) estimate human cognitive states to monitor the human partner, 2) determine the best communication methods to calibrate cognitive states, and 3) allocate functions among humans and robots to optimize team performance. We demonstrate this approach through drone search and rescue (SAR) missions. 

Personalized driving styles (DS) in automated vehicles (AVs) can improve drivers’ trust and acceptance. However, little is known about how drivers identify and communicate their preferred DSs to AVs. Also, most studies have focused on normal driving situations, without considering risky or off-nominal scenarios. This study examined how drivers find their preferred DS by allowing participants to adjust the AV’s driving behavior. A study was conducted whereby drivers experienced one of four predefined DSs, categorized by their level of conservativeness in terms of safety and speed in navigating road obstacles. Participants completed 10 driving trials where they provided feedback to an AV regarding whether to change or maintain its DS. Drivers preferred to maintain the same DS from previous trials and a DS that was neither too aggressive nor too conservative. Findings can help inform the development of adaptive algorithms that preemptively adjust DS without drivers’ explicit inputs. 

As automated vehicles (AVs) become more intelligent, the concept of teaming with a semi-autonomous AV has gained considerable traction. However, few empirical studies exist that quantify the performance of an AV-(human) driver team. To contribute to the existing work in this area, the purpose of this study was to examine how the availability of both a vehicle and driver agent, or the unavailability of one agent, influenced teaming in driving. The vehicle’s individual goal: manage the dynamic driving task. The driver’s individual goal: complete a work-related document-editing task. Together, the vehicle would ask the driver about potential roadway dangers and the driver would respond to the vehicle’s queries. Overall, introducing availability allowed for function allocations where drivers had individual goals separate from driving. Study results provide initial insights into operationalizing the future vision of AVs where safety and driver productivity during commutes are maximized.