We address the problem of flight control in the presence of actuator anomalies. A supervisory control architecture that includes the actions of both a human pilot and an autopilot is proposed to ensure resilient tracking performance in the presence of anomalies. The pilot is tasked with supervisory, higher level decision-making tasks, such as anomaly detection, estimation, and command regulation. The autopilot is assigned a lower level task of accurate command following based on an adaptive control design. The main innovations in the proposed architecture are the use of human pilot in utilizing the concepts of capacity for maneuver (CfM) and graceful command degradation (GCD), both of which originate in cognitive sciences and a judicious combination of the pilot inputs and the autopilot control action. Together, they provide guidelines for a system to be resilient, which corresponds to the system's readiness to respond to unforeseen events. The supervisory control architecture is shown to be capable of achieving maximum CfM while allowing minimal GCD, as well as satisfactory command following postanomaly, resulting in resilient flight capabilities. The proposed controller is analyzed in a simulation study of a nonlinear F-16 aircraft model under actuator anomalies. It is shown through numerical studies that under suitable inputs from the pilot, the overall controller is able to deliver resilient flight.

Previous studies suggest that functional ankle instability (FAI) may be associated with deficits in the ability to sense muscle forces. We tested individuals with FAI to determine if they have reduced ability to control ankle muscle forces, which is a function of force sense. Our test was performed isometrically to minimize the involvement of joint position sense and kinesthesia. A FAI group and a control group were recruited to perform an ankle force control task using a platform-based ankle robot. They were asked to move a cursor to hit 24 targets as accurately and as fast as possible in a virtual maze. The cursor movement was based on the direction and magnitude of the forces applied to the robot. Participants underwent three conditions: pre-test (baseline), practice (skill acquisition), and post-test (post skill acquisition). The force control ability was quantified based on the accuracy performance during the task. The accuracy performance was negatively associated with the collision count of the cursor with the maze wall. The FAI group showed reduced ability to control ankle muscle forces compared to the control group in the pre-test condition, but the difference became non-significant in the post-test condition after practice. The change in performance before and after practice may be due to different degrees of reliance on force sense.

Ankle joint plays a critical role in daily activities involving interactions with environment using force and position control. Neuromechanical dysfunctions (e.g. due to stroke or brain injury), therefore, have a major impact on individuals’ quality of life. The effective design of neurorehabilitation protocols for robotic rehabilitation platforms, relies on understanding the control characteristics of the ankle joint in interaction with external environment using force and position. This is particularly of interest since the findings in upper-limb may not be generalizable to the lower-limb. This study aimed to characterize the skilled performance of ankle joint in visuomotor position and force control. A 2-degree of freedom (DOF) robotic footplate was used to measure individuals’ force and position. Healthy individuals (n = 27) used ankle force or position for point-to-point and tracking control tasks in 1-DOF and 2-DOF virtual game environments. Subjects’ performance was quantified as a function of accuracy and completion time. While the performance measures in 1-DOF control tasks were comparable, the subjects’ performance in 2-DOF tasks was significantly better with position control. Subjective questionnaires on the perceived difficulty matched the objective experimental results; suggesting that the poor performance in force control was not due to experimental setup or fatigue but can be attributed to the different levels of challenge needed in neural control. It is inferred that in visuomotor coordination, the neuromuscular specialization of ankle provides better control over position rather than force. These findings can inform the design of neuro-rehabilitation platforms, selection of effective tasks, and therapeutic protocols.

Ankle sprain is a significant public health concern that can compromise ambulation and activities of daily living. An effective rehabilitation protocol includes objective range of motion (ROM), strength, and proprioception assessment and training. The virtually interfaced robotic ankle and balance trainer (vi-RABT) is a platform-based robot to streamline the ankle and balance rehabilitation. vi-RABT is a 2-degree of freedom robot about dorsiflexion/plantarflexion and inversion/eversion of ankle joint. It has a compact electromechanical design instrumented with actuators, angle and torque sensors and equipped with an impedance controller. vi-RABT hosts interactive games, which are designed by therapists, to turn the repetitive therapy into a more engaging experience. The system was used in a preliminary study for ankle joint assessment and training of two healthy human subjects. The assessment results were compared with outcomes using standard equipment in which the ankle joint ROM and strength were found close to the benchmark measures. In training blocks, the impedance controller corrected the individuals’ motion in the goal-oriented interactive game, improving the movement speed and accuracy while delivering satisfactory torque and angle tracking performance. The preliminary results shows that vi-RABT can streamline ankle joint assessment and training in the seated posture. The relatively smaller size of vi-RABT may increase the range and frequency of applications in private clinics and hospitals.

In this paper, we consider a flight control problem subjected to actuator anomalies. We propose a shared control architecture that includes actions of both a human pilot and an adaptive autopilot that achieves bumpless performance while retaining a Capacity for Maneuver (CfM) following the anomaly. The shared controller distributes the critical tasks of anomaly perception and adaptation between a human pilot and an adaptive autopilot to best utilize their individual strengths. The human pilot’s expertise in detection and perception of anomaly is utilized to determine an initial estimate of the plant parameters. An adaptive autopilot then carries out further tuning of the controller to minimize the tracking error while retaining a certain CfM. The shared controller is evaluated in simulation study of an F-8 aircraft under actuator anomalies and shown to be advantageous compared to a fixed-gain fault tolerant controller.

Lower extremity function recovery is one of the most important goals in stroke rehabilitation. Many paradigms and technologies have been introduced for the lower limb rehabilitation over the past decades, but their outcomes indicate a need to develop a complementary approach. One attempt to accomplish a better functional recovery is to combine bottom-up and top-down approaches by means of brain-computer interfaces (BCIs). In this study, a BCI-controlled robotic mirror therapy system is proposed for lower limb recovery following stroke. An experimental paradigm including four states is introduced to combine robotic training (bottom-up) and mirror therapy (top-down) approaches. A BCI system is presented to classify the electroencephalography (EEG) evidence. In addition, a probabilistic model is presented to assist patients in transition across the experiment states based on their intent. To demonstrate the feasibility of the system, both offline and online analyses are performed for five healthy subjects. The experiment results show a promising performance for the system, with average accuracy of 94% in offline and 75% in online sessions.

While asymmetries have been observed between the dominant and non-dominant legs, it is unclear whether they have different abilities in isometric force control (IFC). The purpose of this study was to compare ankle IFC between the legs. IFC is important for stabilization rather than object manipulation, and people typically use their non-dominant leg for stabilization tasks. Additionally, studies suggested that a limb can better acquire a motor task when the control mechanism of the task is related to what the limb is specialized for. We hypothesized that the non-dominant leg would better (1) control ankle IFC with speed and accuracy, and (2) acquire an ankle IFC skill through direct learning and transfer of learning. Two participant groups practiced an IFC task using either their dominant or non-dominant ankle. In a virtual environment, subjects moved a cursor to hit 24 targets in a maze by adjusting the direction and magnitude of ankle isometric force with speed (measured by the time required to hit all targets or movement time) and accuracy (number of collisions to a maze wall). Both groups demonstrated similar movement time and accuracy between the dominant and non-dominant limbs before practicing the task. After practice, both groups showed improvement in both variables on both the practiced and non-practiced sides (p < .01), but no between-group difference was detected in the degree of improvement on each side. The ability to control and acquire the IFC skill was similar between the legs, which did not support the brain is lateralized for ankle IFC.

A shared flight control framework between an autopilot and a human pilot is proposed to ensure resilient performance following an anomaly. The human pilot is modeled using a perception component and an adaptation component. The concept of Capacity for Maneuver (CfM) is used to develop the perception component, and an adaptation component similar to the concepts proposed in the flight control literature is used. The shared control architecture is evaluated in the context of flight control, where an anomaly is modeled as a sudden change in the underlying flight dynamics, with resilient performance defined as reduced command tracking error. Simulation studies show that with no shared control, introduction of an anomaly significantly degrades the resilience, while the proposed shared control results in almost identical performance after the anomaly. Proceedings of the International Symposium on Sustainable Systems and Technologies (ISSN 2329-9169) is published annually by the Sustainable Conoscente Network. Jun-Ki Choi and Annick Anctil, co-editors 2016. ISSSTNetwork@gmail.com. Copyright © 2016 by Amir B. Farjadian, Anuradha M. Annaswamy, David Woods Licensed under CC-BY 3.0. Cite as: Towards A Resilient Control Architecture: A Demonstration of Bumpless Re-Engagement Following an Anomaly in Flight Control. Proc. ISSST, Amir B. Farjadian, Anuradha M. Annaswamy, David Woods. Doi information v4 (2016) Proceedings of the International Symposium on Sustainable Systems and Technologies, v4 (2016) A. B. Farjadian, et al. Introduction. Enhancing automated control capabilities can improve precision, speed, and robustness to well modeled disturbances. Nevertheless, automated processes have limits that define a boundary or envelope. When conditions, context, and disturbances occur that fall outside of this envelope, surprising events can occur and produce a cascade of additional disturbances that exceed the capabilities of automated control (Woods and Sarter 2000). Such anomalies require engagement of human supervisors from other activities to re-assess the situation and intervene quickly and decisively to forestall failures — a form of shared control that can be termed as Bumpless Re-engagement. But today when these human supervisory functions are needed, they are poorly supported, cost intensive, and often slow or erroneous (Woods and Hollnagel 2006). In other words, shared control breaks down, and a shift is required from “textbook” autonomous performance to handling anomalous events with high potential to cascade toward failure. Current forms of shared control assume a partially autonomous machine does all of the work to handle variability — until external demands imposed on the machine exceed the automation's capabilities to handle the situation — then control is transferred to people who have to take over when the situation is difficult to handle. This form of shared control virtually guarantees bumpy and late transfers of control that increase the risk of decompensation — inability of a humanmachine system to keep pace with growing or cascading demands (Woods 2011). In real cases of human supervision of automation based on this model, the bumpy and late transfers of control have contributed to actual accidents (Job 1998; Woods 2006, Chapter 10). This paper uses a classic flight control problem to introduce a novel approach to provide improved shared control. The new method is based on the concept of Capacity for Maneuver (CfM) – the remaining range or capacity to continue to respond to ongoing and upcoming demands (Woods and Branlat 2011). Control then should seek to minimize the risk of exhausting a unit’s capacity for maneuver as that agent responds to changing and increasing demands (risk of saturation). We are proposing a resilient shared control architecture based on reducing the risk of saturating CfM that allows a timely and effective human re-engagement following an anomaly to sustain a desired tracking performance. A problem in flight dynamics is chosen (Hess 2015) to demonstrate the potential benefits of the new architecture as it can involve human-machine interaction in response to anomalous situations (Belcastro et al. 2010; Glussich et al. 2010; Woods and Sarter 2000). Control performance is compared in the case of fully automated flight control with the new shared control architecture, where a model of the human pilot takes over control following an anomaly. The human pilot model uses the new resilient control approach utilizing the information about CfM. Flight Control Problem and Shared Control. Bumpy transfer of control to human pilot following an anomaly can result in loss of control and disastrous consequences. Figure 1 shows a simple block diagram of manual/auto-pilot control of an aircraft, where Yc(s) denotes the flight dynamics, M denotes a flight variable of interest (e.g. flight path angle, or roll angle), and the goal is for M to follow the command signal as closely as possible. GM(s) and GA(s) denote the mathematical model of the manual pilot and the auto-pilot, respectively, representing the response time and gain characteristics of the two controllers, with s denoting the differential operator d/dt. The goal is the timely and effective switching between GM(s) and GA(s) so as to minimize the error between M and the command signal (Mcmd), especially after the occurrence of an anomaly. A. B. Farjadian, et al. Figure 1: Shared Aircraft Control. The aircraft is controlled by autopilot in nominal case. The pilot takes over control following any special occurrences or anomalies. Flight control is a well understood and well researched topic in aeronautics, with several textbooks written on the subject (Stevens et al. 2015; and Lavretsky et al. 2012). Basic principles of autopilot design that ensures satisfactory command tracking are laid out in these texts and elsewhere in the literature. Mathematical models of human pilot behavior have also been researched extensively, with seminal contributions by McRuer (McRuer 1980). Detailed models of the pilot behavior, especially at frequencies where stable feedback action is most urgently needed, have been developed (McRuer 1980; Hess 2006, 2009, 2014, 2015). The result is the cross-over model which has been used extensively in autopilot and fly-by-wire designs for nominal flight control. More recently, these models have been studied (Hess 2009, 2014, 2015) to understand a pilot’s actions following anomalous events that may produce a significant change in the flight dynamics. The framework we develop in this paper builds on autopilot designs and human pilot models that have been proposed in these earlier works (Hess 2015), and consists of a perception component, which detects the occurrence of an anomaly in a swift and accurate manner, and an adaptation component that prompts the pilot to improve performance. The perception component is built on the notion of resilient control using the actuator’s capacity for maneuver (CfM). Based on the results of the perception metric, the adaptation component seeks to sustain the tracking error within acceptable limits in the shortest time window. The overall resilient control architecture we propose consists of both the perception and adaptation components. The paper presents results that, with this architecture, the tracking performance remains within the desired boundary even after the occurrence of an anomaly. Shared Control Architecture. A schematic of the autopilot and the manual pilot action for flight control are shown in Figures 2(A) and 2(B). In this figure, Yc(s) denotes the aircraft dynamics. E and R denote the perceived position and rate error respectively. The autopilot action consists of measuring a rate ?̇? and a position M, and using those measurements together with the command and feedback control gains Kr and Kp to compute the necessary compensation, denoted as v. This signal is in turn fed to an actuator in the aircraft, which is a transducer that converts this electronic signal into a desired control surface deflection. Every transducer has physical limits and maximum boundaries which is modeled here by a saturation function (f) mainly affecting the amplitude. The values beyond certain threshold (u0) will be clamped as they are not effective on the aircraft actuators. The goal of the overall autopilot design, as mentioned earlier, is to design Kr and Kp to achieve desired tracking performance. While the schematic presented in this figure is somewhat simplified, it encapsulates the general principle of an autopilot. Yc(s) Aircraft GM(s) Manual Pilot GA(s) Autopilot Tc M M d A. B. Farjadian, et al. Yc(s) Kp Kr Aircraft R v u f(.)

Inspired by achievements in rehabilitation, motor learning, and neuroscience, therapeutic robots are aiming to provoke neuromotor plasticity and improve recovery after stroke and mobility impairments. Human sensorimotor system is specialized with position and velocity sensory fibers and exhibits variant muscle impedance in accordance with the ongoing task. The virtually interfaced robotic ankle and balance trainer (vi-RABT) was introduced as a cost-effective platform-based rehabilitation robot to improve overall ankle / balance strength, mobility and control. This study is the first step toward assistive / resistive ankle rehabilitation using vi-RABT. We have implemented a task-dependent anisotropic impedance controller into the 2-DOF ankle rehabilitation robot. An objective virtual Maze game is developed. The controller is specifically designed for the Maze workspace; and exhibits elastic or pure viscos properties in compliant with the subject's direction of movement. Early results on two human subjects are presented.

Each year in the US, 628,000 people suffer an ankle sprain, and 795,000 suffer a new or recurrent stroke. Due to improved survival rates after stroke, significant increases in stroke population are projected by 2030. So far, there is no cost-effective robotic ankle/balance trainer in the market. In this paper, we present the Virtually-Interfaced Robotic Ankle and Balance Trainer (vi-RABT), a low-cost robotic system that will improve overall ankle / balance strength, mobility and control. The system is equipped with 2 degrees of freedom (DOF) controlled actuation along with complete means of force and angular measurements. The preliminary results on a single robotic footplate confirm the system design. The system will be used for measurement of ankle kinematics, ankle kinetics and balance function, as well as for retraining motor control and strength of the ankle during plantarflexion / dorsiflexion (PF/DF), ankle inversion / eversion (IN/EV) and circumduction motions.

The ankle joint is critical to upright balance control and gait function, and is a common location for orthopedic or neurological injuries - e.g., every year in the US 628,000 people suffer ankle sprains and 795,000 suffer a stroke. The Virtually-Interfaced Robotic Ankle and Balance Trainer (vi-RABT) is a rehabilitation system designed to help such individuals improve their ankle and balance control. The system is equipped with two actuators and complete mechanisms for torque and angle measurement. A model-based proportional-integral-derivative (PID) controller was tuned and implemented into the system and an interactive maze game was augmented to provide an entertaining rehabilitation experience. We present results of a pilot test on these system features. Vi-RABT shows promise for treating a wide variety of ankle and balance disorders.

An estimated of 2,000,000 acute ankle sprains occur annually in the United States. Furthermore, ankle disabilities are caused by neurological impairments such as traumatic brain injury, cerebral palsy and stroke. The virtually interfaced robotic ankle and balance trainer (vi-RABT) was introduced as a cost-effective platform-based rehabilitation robot to improve overall ankle/balance strength, mobility and control. The system is equipped with 2 degrees of freedom (2-DOF) controlled actuation along with complete means of angle and torque measurement mechanisms. Vi-RABT was used to assess ankle strength, flexibility and motor control in healthy human subjects, while playing interactive virtual reality games on the screen. The results suggest that in the task with 2-DOF, subjects have better control over ankle's position vs. force.

Ankle impairment and lower limb asymmetries in strength and coordination are common symptoms for individuals with selected musculoskeletal and neurological impairments. The virtual reality augmented cycling kit (VRACK) was designed as a compact mechatronics system for lower limb and mobility rehabilitation. The system measures interaction forces and cardiac activity during cycling in a virtual environment. The kinematics measurement was added to the system. Due to the constrained problem definition, the combination of inertial measurement unit (IMU) and Kalman filtering was recruited to compute the optimal pedal angular displacement during dynamic cycling exercise. Using a novel benchmarking method the accuracy of IMU-based kinematics measurement was evaluated. Relatively accurate angular measurements were achieved. The enhanced VRACK system can serve as a rehabilitation device to monitor biomechanical and physiological variables during cycling on a stationary bike.

Electrical onset of an epileptic seizure is characterized by low frequency and high amplitude rapid discharges at hippocampus. An intelligent controller based on emotional learning algorithm of the brain has been developed to abate this bursting activity. The control input has been applied to the lumped parameter model of epilepsy with the purpose of steering the epileptic spikes to normal activity. The results reveal that rapid discharges occurring at seizure onset can be manipulated by applying bounded stimuli to the model.

Sliding mode is an acknowledged nonlinear robust control method which suffers from chattering phenomenon, the destructive high-frequency oscillations in internal states and control signal. One of the suggested routines to reduce chattering is to replace the discontinuity switching term, in standard method formulation, with a saturation function. Considering the fact that saturation function, with fixed-gradient, reduces the performance; we utilize an adaptive-gradient saturation function to overcome this limitation. Reinforcement learning algorithm is employed to find the instantaneous optimal value for the gradient of saturation function, with the ultimate goal of chattering reduction. The proposed intelligent sliding mode controller is applied to the tracking problem of chaotic Lorenz plant whereas the agent is rewarded (punished) for lower (higher) chattering. Simulation results are reported for standard and intelligent sliding mode controllers. The efficient control signal density as well as lower tracking error was attained after the agent learned the dynamics of the complex chaotic plant. Incorporating reinforcement learning into robust nonlinear control theory shows a promising route to achieve better performance.

Stroke is a leading cause of serious long-term disability in the United States. There is a need for new technological adjuncts to expedite patients' scheduled discharge from hospital and pursue rehabilitation procedure at home. SQUID is a low-cost, smart shirt that incorporates a six-channel electromyography (EMG) and heart rate data acquisition module to deliver objective audiovisual and haptic biofeedback to the patient. The sensorized shirt is interfaced with a smartphone application, for the subject's usage at home, as well as the online database, for the therapist's remote supervision from hospital. A single healthy subject was recruited to investigate the system functionality during improperly performed exercise. The system can potentially be used in automated, remote monitoring of variety of physical therapy exercises, rooted in strength or coordination training of specific muscle groups.

Electrical stimulation is a less-invasive alternative for treating drug-resistant epilepsy compared to surgical resection of the epileptogenic area. Different stimulation protocols have been practiced to suppress seizures either in-vivo, in-vitro and in-silico. In this work we have controlled rapid discharges occurring at electrical onset of seizure in an in-silico model of epilepsy using backstepping technique.