Advances in Human Factors in Robots and Unmanned Systems

To realize the full benefit from autonomy, systems will have to react to unknown events and uncertain dynamic environments. The resulting number of behaviors is essentially infinite; thus, the system is effectively non-deterministic but an operator needs to understand and trust the actions of the autonomous vehicles. This research began to tackle non-deterministic systems and trust by beginning to develop a user trust function based on intent information displayed and the prescribed bounds on allowable behaviors/actions of the non-deterministic system. Linear regression shows promise on being able to predict a person’s confidence of the machine’s prediction. Linear regression techniques indicated that subject characteristics, scenario difficulty, the experience with the system, and confidence earlier in the scenario account for approximately 60% of the variation in confidence ratings. This paper details the specifics of the liner regression model – essentially a trust function – for predicting a person’s confidence.

The purpose of this investigation was to assess user preferences for controlling an autonomous system. A comparison using a virtual environment (VE) was made between a joystick based, game controller and a gesture-based system using the leap motion controller. Command functions included basic flight maneuvers and switching between the operator and drone view. Comparisons were made between the control approaches using a representative quadcopter drone. The VE was designed to minimize the cognitive loading and focus on the flight control. It is a physics-based flight simulator built in Unity3D. Participants first spend time familiarizing themselves with the basic controls and vehicle response to command inputs. They then engaged in search missions. Data was gathered on time spent performing tasks, and post test interviews were conducted to uncover user preferences. Results indicate that while the gesture-based system has some benefits the joystick control is still preferred.

This work addresses the factor of degraded automation reliability of machine based aerial reconnaissance in a manned-unmanned teaming approach. An army transport helicopter is accompanied by three unmanned aerial vehicles for reconnaissance purposes, guided by the helicopters crew. Automated capabilities onboard the UAVs offer high automated, task-based guidance as well as manual operation. We designed and implemented an assistance system in our helicopter flight simulator, that supports the commander in gaining relevant reconnaissance information on flight routes for the helicopter to follow. Due to imperfection in automated reconnaissance performed by machine algorithms, we explicitly regarded the aspect of degrading reliability by utilizing the paradigm of “Levels of Automation”. The automation system produces reconnaissance results, thereby considering differing automation reliability. Several data representation modes were applied to display preprocessed results in the helicopters multi-function displays. We conducted an extensive human-in-the-loop campaign with army helicopter crews in full mission scenarios, in which system-triggered changes between the automation levels occurred and the cooperative human-machine relationship changed online. This paper presents questionnaire-gathered results of our investigation during mission execution, shedding light on human factors, user acceptance and system design aspects.

Autonomous ground vehicles (AGVs) play a significant role in performing the versatile task of replacing human-operated vehicles and improving vehicular traffic. This facilitates the advancement of an independent and interdependent decision-making process that increases the accessibility of transportation by reducing accidents and congestion. Presently, human-machine cooperation has focused on developing advanced algorithms for intelligent path planning and execution that is functional in providing reliable transportation. From industry simulations to field tests, AGVs exhibited various mishaps or errors that have a probability to cause fatalities and undermine the potential benefits. Therefore, it is very important to focus on reducing fatalities due to either human error or AGV system error. To solve this problem, the paper proposes an error prediction model to reduce AGV errors through appropriate human intervention. In this paper, we use the data from AGV exteroceptive sensors such as stereo-vision cameras, long and short range RADARS, and LiDAR to predict the AGVs error through Dempster–Shafer theory (DST) based on sensor data fusion technique. The results obtained in this work suggest that there is a lot of scope for improvement in the performance of AGV when conflicts are predicted in advance and alerting human for intervention. This would, in turn, improve human-machine cooperation.

Advanced autonomous maritime operations are today an emerging academic field, where the implementation of autonomous or semi-autonomous control, support and maintenance systems. The semi-autonomous operations often require a complex interaction between human knowledge and experience as well as suitable intelligent based programs. In this simulated approach of a ship’s berthing operation, the captains’ experience and knowledge is the basis for training the fuzzy logic system. The human-machine interaction can further be enhanced by a second fuzzy logic system to feedback the out-put fuzzy logic signal and adjust the berthing maneuver to find near-optimal solutions. The paper will present an Artificial Intelligent based semi-autonomous solution in maritime operations and discuss the related human factors as well as the sensors needed to define the decision support system for ship berthing operation and demonstrating by a proposed fuzzy logic-based solution.

Paper presents unmanned aerial vehicle (UAV) usage options and importance to fulfil power engineering tasks, improvements and their explanation. This paper also is part from research in this field restarted this year based also on previous research connected with power consumption problems [1] and propose initially data according topology without full coverage of data necessary to present final data.

In this paper we discuss how team configuration may influence how information is shared among team members for low-altitude Unmanned Aircraft Systems (UAS) operations. NASA collected and analyzed observation data gathered during a series of field tests for the UAS Traffic Management (UTM) project. The field tests were part of a larger effort aimed at advancing the UTM concept, conducted at six test-sites across the USA. Ground control station (GCS) concepts, flight-crew composition, and crew-size varied within and across test-sites. Flight crews took two strategic approaches to organizing their teams. The first of the two approaches was implemented by one third of the flight crews. These crews integrated the role of UTM operator into the duties of existing crew members, merging the current roles with this new one, keeping the UTM operator collocated with the flight crew. The remaining two thirds implemented a distributed team configuration, where a single UTM operator distributed support across multiple crews. Results from our data collection efforts revealed that UTM operator location influenced whether flight crews used verbal communication versus displays to acquire UTM information.

This paper presents a novel teleoperation interface for continuous backbone continuum robots. Previous teleoperation interface methods for continuum robots were less intuitive due to a degree-of-freedom mismatch, using non-continuum interface input devices with fewer degrees-of-freedom than the robot that was being operated. The approach introduced in this paper uses a graphical 3D model on screen to directly operate the continuum robot for an easier user experience. This paper details the development of both the model and software. The teleoperation interface was developed specifically for a nine degree-of-freedom pneumatically driven extensible continuum robot, but it can easily be extended to any continuum robot with an arbitrary number of section due to its modular design. Experiments using the aforementioned system on two different continuum robots are reported and areas for future work and improvement are detailed.

Autonomous tools that can evaluate a course of action (COA) are being developed to assist military leaders. System designers must determine the most effective method of presenting these COAs to operators. To address this challenge, an experimental testbed was developed in which participants were required to achieve the highest score possible in a specific time window by completing mission tasks. For each task, eight possible COAs were presented. Each COA had four parameters—points, time, fuel, and detection. Four experimental visualizations were evaluated, varying in COA number and type: (1) a single COA (most points), (2) four COAs (four highest point values), (3) four COAs (the most points, the least time, the least fuel, and the least chance of detection), and (4) all eight COAs. Both objective and subjective data indicated that the single COA visualization was significantly less effective than the other visualizations. Suggestions are made for follow-on research.

Trust is paramount to the development of effective human-robot teaming. It becomes even more important as robotic systems evolve to make both independent and interdependent decisions in high-risk, dynamic environments. Yet, despite decades of research looking at trust in human-interpersonal teams, human-animal teams, and human-automation interaction, there are still a number of critical research gaps related to human-robot trust. The US Army Research Laboratory Robotics Collaborative Technology Alliance (RCTA) is a 10-year program with government, industry and academia combining to conduct collaborative research across four major robotic technical areas of intelligence, perception, human-robot interaction, and manipulation and mobility. This paper describes findings from over 60 publications and 49 presentations describing research conducted as part of the RCTA from 2010 to 2017 to address these critical gaps on human-robot trust.

We present a stability analysis of the upright stance of a model of a humanoid robot balancing on a rotating board and driven by a human-inspired control strategy. The humanoid-board system is modeled as a triple inverted pendulum actuated by torques at the board’s hinge, ankle joint, and hip joint. The ankle and hip torques consider proprioceptive and vestibular angular information and are affected by time delays. The stability regions in different parameter’ spaces are bounded by pitchfork and Hopf’s bifurcations. It is shown that increasing time delays do not affect the pitchfork but they shrink the Hopf bifurcations. Moreover, the human-inspired control strategy is able to control the upright stance of a humanoid robot in the presence of time delays. However, more theoretical and experimental studies are necessary to validate the present results.

In this publication, we investigate how the term ergonomics could be defined for human-robot collaboration (HRC) as a sociotechnical system (STS). Thus, we compare different definitions of ergonomics and human factors and conclude on a definition suggested for adoption. Moreover, we compile a list of human factors relevant to that context. We conducted this investigation, because HRC is mainly viewed from a technical viewpoint, although the implications of human involvement should not be underestimated. However, ergonomic evaluation of HRC is based on old methods. The main purpose of this publication is therefore to contribute to a foundation for new ergonomic evaluation methods.

It is increasingly common for people to work alongside robots in a variety of situations. When a robot is completing a task, the human operator of the robot may be present at the work site or operating remotely. It is important to understand how people’s behavior towards a robot is influenced by the presence or absence of the operator. We observed individuals in public locations as they pass by a robot with and without a visible human operator. We show that individuals were more likely to approach and interact with the robot when it was alone. Also, individuals that interacted with the robot when it was alone were more likely to take extra candy from the robot. Our results suggest that robots with visible operators discourage interaction with the robot but encourage honesty in interactions with the robot.

Robotics, combined with computer science and human-centered studies, can have a substantial impact in areas such as education and innovation. Robots have proven to be a good tool to gain and maintain users’ involvement in different activities. In education, robots can be used as teaching assistants to improve participation, enhance concentration or just to get students’ attention. In this research, we involved an NAO, a humanoid robot, in a workshop presentation with the aim of measuring the impact of this technique on the level of engagement showed by the participants. The robot was programmed to simulate speech and gesticulate while it talked to apply the Wizard of Oz technique.

A motion planner based on artificial potential field for unmanned tractor is proposed in this paper. In order to get the effective environment model, the model for condition of terrain and impact of ground to tractor is analyzed and built. The tractor usually brings a trailer behind, so the kinematics of tractor with a trailer is presented. The motion planner based on artificial potential field is designed for the unmanned tractor working in farmland. According to the characteristics of the unmanned tractor, the control algorithm and motion planner is optimized. The simulation of the improved motion planner for unmanned tractor is presented and analyzed. And the simulation results are presented to show the effectiveness of the proposed method.

A non-humanoid robot is used to assist in an educational workshop of Quality Assurance and DevOps. The goal of this research was to determine the level of engagement shown by students of computer science in a presentation conducted by a University professor and assisted by a robot. The robot interaction was based on the Wizard of Oz technique. The order of actions between the professor and the robot was scripted and practiced before the workshop. After the workshop, a survey was conducted to assess the students’ perception towards robot’s shape, size, behavior and, performance. The survey also included the Godspeed Questionnaire Series to measure participant’s perception of the robot and its effectiveness as an educational assistant. The results revealed the participants considered the robot featured personalized cognitive skills and exhibited an acceptable integration in the workshop.