Keywords: Shared control, Semi-autonomous and mixed-initiative systems, Potential impact of automation and open problems
Abstract: People tend to overtrust sophisticated computing devices, especially those powered by AI. As these systems become more fully interactive with humans during the performance of day-to-day activities, ethical considerations in deploying these systems must be more carefully investigated. Bias, for example, has often been encoded in and can manifest itself through AI algorithms, which humans then take guidance from, resulting in the phenomenon of excessive trust. Bias further impacts this potential risk for trust, or overtrust, in that these cyber-physical systems are learning by mimicking our own thinking processes, inheriting our own implicit gender and racial biases, for example. These types of human-AI feedback loops may consequently have a direct impact on the overall quality of the interaction between humans and machines, whether the interaction is in the domains of healthcare, job-placement, or other high-impact life scenarios. In this talk, we will discuss various forms of bias, as embedded in our machines, and possible ways to mitigate its impact on cyber-physical human systems.

Keywords: Shared control
Abstract: Electromyography (EMG) interfaces are a promising alternative to traditional manual interfaces such as joysticks, mice, and touchscreens for applications such as prosthetics, rehabilitation, and human-computer interaction. McRuer's crossover model has been extensively studied to determine the impacts of dynamical systems on humans using manual interfaces; however, the same analysis has not been conducted with EMG interfaces or more complex dynamical systems. In this paper, we establish and assess changes in human parameters (gain and delay) and bandwidth for manual (joystick) and EMG interfaces when humans are tasked with controlling a first- and second-order dynamical system. We performed a secondary data analysis to estimate the human parameters for 11 participants by performing least-squares fitting on the error between empirical estimates (calculated from measured signals and system dynamics at specific frequencies) and parameterized models (developed from the McRuer’s gain-margin crossover model). EMG delay was smaller than the manual delay for the first-order system and EMG delay was smaller with the first-order system than the second-order system. EMG bandwidth was also larger than the manual bandwidth for both first- and second-order systems. These results suggest that using an EMG interface improves the user’s reaction time in a first-order system, and the EMG interface increases the bandwidth that the user can control for both first- and second-order systems compared to a manual interface. Understanding the differences in delays and bandwidth based on interfaces and system dynamics is useful for designing multimodal interfaces or for complex systems where the human delay or bandwidth is important.

Keywords: Intelligent road transportation, Shared control
Abstract: The prevalence of Level 2 vehicle automation on U.S. roadways is increasing. As such, drivers are responsible for monitoring the automation and taking over control as necessary. However, it remains unclear when a driver may begin to exhibit behavioral responses that could indicate their intention to takeover. In this paper, we use an exhaustive approach to determine the features that best predict takeover, along with the time windows over which those features should be sampled. Specifically, we consider features that can be measured in real time and that are predominantly driver-based, including both behavioral and psychophysiological features. The resulting analysis highlights pupil diameter as the most significant predictor of takeover behavior. Finally, investigation into feature extraction windows indicates that window size may be feature-specific and may not generalize across features of the same modality. These results have significance for what types of sensors should be chosen for takeover prediction in L2 automated vehicles in which real-time takeover prediction is of interest.

Keywords: Decision-support for human operators, Potential impact of automation and open problems, Automotive cooperated control (ADAS, etc)
Abstract: Variability in human response creates non-trivial challenges for modeling and control of human-automation systems. As autonomy becomes pervasive, methods that can accommodate human variability will become paramount, to ensure efficiency, safety, and high levels of performance. We propose an easily computable modeling framework which takes advantage of a metric to assess variability in individual human response in a dynamic task that participants repeat over several trials. Our approach is based in a transformation of observed trajectories to a reproducing kernel Hilbert space, which captures variability in human response as a distribution embedded within the Hilbert space. We evaluate the similarity across responses via the maximum mean discrepancy, which measures the distance between distributions within the Hilbert space. We apply this metric to a difficult driving task designed to elucidate differences across participants. We conducted a pilot study with 6 participants in an advanced driving simulator, in which participants were tasked with collision avoidance of an obstacle in the middle of the road, around a blind corner, in a nighttime scenario, while steering only with the non-dominant hand.

Keywords: Smart cities
Abstract: In the context of sustainable lifestyle it has been observed that, while expressing eco-positive attitudes, individuals often do not act accordingly in their habitual behavior. This gap, termed the ``value-action'' gap, has been explained in terms of desire to seek social approval or as a consequence of the presence of overriding conflicting goals, associated for instance with material costs. In this work, we study a two-scale networked model for dynamic decision-making in which interacting agents are able to exchange opinions and discuss the different reasons they produce their choices and, in addition, are able to observe the actions of their neighbors in the network and adjust their preferences. Coupling on the two scales leads to a reduced value-action gap, and ultimately to a consensus. A numerical example illustrates the effect that tradeoffs between goals and social pressure have on the behavior of the group.

Keywords: Shared control, Semi-autonomous and mixed-initiative systems, Potential impact of automation and open problems
Abstract: Jordan Berg and Alex Leonessa of the NSF Directorate for Engineering (ENG) will present funding opportunities for research projects on human-machine interactions, focusing on the Dynamics, Control, and Cognition (DCC) cluster in the ENG Division of Civil, Mechanical, and Industrial Innovation (CMMI) . There will be a brief description of the DCC programs, namely Dynamics, Control, and System Diagnostics (DCSD), Foundational Research in Robotics (FRR), and Mind, Machine, and Motor Nexus (M3X). DCSD supports fundamental advances in dynamics and control, including dynamics arising in human-machine interactions and techniques used to shape those dynamics. FRR supports research to create robots with new or substantially improved capabilities, including capabilities for human interaction. M3X focuses on interactions between humans and intelligent machines that are primarily mediated through physical forces or otherwise non-symbolic channels. Half the session will be reserved for Q&A.

Keywords: Smart Grid and Demand Response, Urban mobility, Smart cities
Abstract: Electric vehicles (EVs) are the key driver to coordinate the sector-coupled system for energy and mobility. Meanwhile, the mobility operation of EVs depends on the consumer behavior of the EV owner. This paper considers a dynamic energy and mobility allocation with the EV consumer behavior in the dynamical sector-coupled system to stabilize a transmission power network and optimize a traffic flow management via EVs. After formulating the dynamical systems coupling the transmission power systems and the traffic systems, we propose a dynamic optimal energy/mobility allocation problem to minimize the physical cost and the behavior cost during the planned period. To analyze the sharing of EVs in the aspect of both energy and mobility, we illustrate and discuss the effectiveness and the implementation of our proposed allocation through simulation imitating the Tokyo area.

Keywords: Decision-support for human operators, Shared control, Intelligent road transportation
Abstract: This paper addresses light guidance control for human-driven vehicles. We manipulate pace-making-light (PML) installed in the road environment to make drivers accelerate their vehicles unconsciously. Also, since the performance of the control system relies on the human reaction, we address the modeling of the driver’s behavior guided by PML. To this end, we tested human subjects by using a driving simulator developed in a virtual reality environment. The collected data were used to model the driver’s behavior and to perform a PML control simulation. In particular, the driver model is utilized for the design of a model predictive controller that is implemented as the PML logic.

Keywords: Remote operation of robotic teams
Abstract: In this paper, we investigate a scenario of one-human-multiple-robot navigation in three dimensions, and examine the impacts of the VR (Virtual Reality) technology on human properties from a control-theoretic perspective. We start by reviewing a passivity-based distributed control architecture that takes complementary interactions such that motion synchronization is autonomously completed by a distributed robot controller while the operator is dedicated to robot navigation. Due to the limited human capability of 3-D recognition and limited dimensionality on the real-time manipulability, 3-D navigation is completely different from that of the one- or two-dimensional case and we need to carefully design both feedback and command interfaces between the operator and robots. Specifically, we employ two different pairs of the interfaces, traditional joystick controller with 2-D display monitor and HMD (Head Mounted Display) with VR controller. We then build human models from the operation data with these interfaces on a human-in-the-loop simulator. Through the human modeling, we present two novel findings: (i) VR interfaces improve the accuracy of the human model with about 25∼45% of fitting ratio, which must drastically eases the design of human-robot collaboration systems, (ii) VR interfaces enhance human passivity, which is a key to ensuring closed-loop stability for the human-in-the-loop system.

Keywords: Assistive devices, Decision-support for human operators
Abstract: The safety guarantee of a control system under input disturbances is challenging. In this paper, we introduce an input-to-state constrained safety zeroing control barrier function (ISCSf-ZCBF) and propose an ISCSf-ZCBF-based human assist controller to ensure the safety of a human-operated system with unknown bounded input disturbances. The proposed human assist controller ensures not only safety but continuity and minimal intervention, which are suitable properties for human assist control. Then, we employ an ISCSf-ZCBF to achieve time-varying obstacle avoidance for an electric wheelchair by regarding an estimation error of obstacle velocity as an input disturbance. Finally, we experimentally confirm the effectiveness of the proposed controller.

Keywords: Urban mobility, Smart cities, Smart Grid and Demand Response
Abstract: In a one-way trip car sharing service, users can return vehicles to a station other than the station from which they departed. While this is highly convenient for users, it also leads to an uneven distribution of the number of vehicles at each station. To cope with this, the service operator needs to relocate the vehicles, but the increase in personnel costs for the relocation makes it difficult to operate the service. In addition, when introducing electric vehicles, the remaining battery capacity and the cost of charging electricity must also be taken into account. Therefore, we propose a pricing system that can verify the business feasibility of the car sharing service by taking these factors into account. The proposed system uses a mathematical model to minimize operational costs and a multinomial logit model to predict demand, and then, determines the monopoly price from an economic perspective based on the costs and demand calculated from these models. This study examines the impact on prices and profits by varying the number of electric vehicles in the region where the car sharing system is introduced.

Keywords: Advanced control design-linear, non-linear, stochastic, large scale control systems, Assistive devices, Shared control
Abstract: Time series anomaly detection is an important process for system monitoring and model switching, among other applications in cyber-physical systems. In this document we present a fast subspace method for time series anomaly detection, with a relatively low computational cost, that has been designed for anomaly detection in real sensor signals corresponding to dynamical systems. We also present some general results corresponding to the theoretical foundations of our method, together with a prototypical algorithm for time series anomaly detection. Some numerical examples corresponding to applications of the prototypical algorithm are presented, and some computational tools based on the theory and algorithms presented in this paper, are provided.

Keywords: Decision-support for human operators, Smart infrastructure, Smart cities
Abstract: In this paper, we analyze the problem of optimally allocating resources in a distributed and privacy-preserving manner. We propose a novel distributed optimal resource allocation algorithm with privacy-preserving guarantees, which operates over a directed communication network. Our algorithm converges in finite time and allows each node to process and transmit quantized messages. Our algorithm utilizes a distributed quantized average consensus strategy combined with a privacy-preserving mechanism. We show that the algorithm converges in finite-time, and we prove that, under specific conditions on the network topology, nodes are able to preserve the privacy of their initial state. Finally, to illustrate the results, we consider an example where test kits need to be optimally allocated proportionally to the number of infections in a region. It is shown that the proposed privacy-preserving resource allocation algorithm performs well with an appropriate convergence rate under privacy guarantees.

Keywords: Comfort control in homes, Advanced control design-linear, non-linear, stochastic, large scale control systems, Connected buildings
Abstract: We present a solution for modeling and online identification for heating, ventilation, and air conditioning (HVAC) control in buildings. Our approach comprises: (a) a resistance-capacitance (RC) model based on first order energy balance for deriving the zone temperature dynamics, and (b) a neural network for modeling HVAC dynamics. State estimation and model identification are simultaneously performed using nonlinear moving horizon estimation (MHE) with physical constraints for system states. We leverage the identified model in model predictive control (MPC) for occupant comfort satisfaction and HVAC energy savings and verify the approach using simulations. Our system relies only on building management system data, does not require extensive data storage, and does not require a detailed building model. This can significantly aid the large scale adoption of MPC for future occupant-centric control of grid-interactive buildings.

Keywords: Exoskeletons, Assistive robotics, Shared control
Abstract: Hybrid exoskeletons are a technology used by people with neurological injuries (e.g., spinal cord injury) for rehabilitation. Hybrid exoskeletons are a hallmark example of physical human-robot interaction because they combine functional electrical stimulation (FES) with a powered exoskeleton. Due to the physical contact between the human and the robot, these exoskeletons must be well controlled to regulate the physical interaction that occurs. In this paper, the exoskeleton is actuated with an integral concurrent learning (ICL) controller to regulate an admittance error system and the human is actuated (via FES) with a robust controller to regulate a position error system. By separating the human dynamics from the hybrid exoskeleton dynamics via the interaction torque, ICL can address structured uncertainties in the exoskeleton dynamics. A Lyapunov-based stability analysis is conducted to prove that the admittance error system is globally exponentially stable under finite excitation. Moreover, the admittance controller is able to ensure bounded position errors, regardless of the FES (i.e., position) controller. A passivity analysis is leveraged to show that the position error system is output feedback passive with respect to the interaction torque, yet is bounded due to the admittance controller.

Keywords: Automotive cooperated control (ADAS, etc)
Abstract: Autonomous Vehicles (AVs) such as cars and trucks are being developed and tested as Cyber-Physical Human systems while the technology improves. Before these systems can achieve full autonomy, some serve as tools in the form of adaptive cruise control.

The CIRCLES Consortium investigates the potential for AVs to increase fuel efficiency of highway traffic by smoothing ``stop-and-go'' traffic waves that result from normal human driving behavior in congestion. We have performed an experiment to evaluate the real world effects of implementing this strategy. A small-scale experiment was performed on I-24 near Nashville, TN in August 2021. This was a precursor to a larger experiment that will take place November 2022. We examine how the human part of the experiment will change as we scale up from an 11 vehicle test (four AVs) to 100 AVs. There are many solutions to problems of the small-scale experiment that would be inconvenient, complicate the experience, or not be practicable.

The small-scale experiment involved eleven cars of which four had a custom control algorithm installed to be engaged by the driver. The large-scale experiment will have 100 cars, all with custom control algorithm installed to act on traffic when the controller is engaged.

We examine key choices made for the small experiment, and how some will be different for the large experiment. Our experience performing the small-scale experiment has made it clear that repeating our methods from this smaller one are ineffective, inefficient, impossible, or stressful for participants if used for the large-scale experiment.