Recent applications of autonomous agents and robots have brought attention to crucial trust-related challenges associated with the current generation ofartificial intelligence (AI)systems. AI systems based on the connectionist deep learning neural network approach lack capabilities of explaining their decisions and actions to others, despite their great successes. Without symbolic interpretation capabilities, they are ‘black boxes’, which renders their choices or actions opaque, making it difficult to trust them in safety-critical applications. The recent stance on the explainability of AI systems has witnessed several approaches toeXplainable Artificial Intelligence (XAI); however, most of the studies have focused on data-driven XAI systems applied in computational sciences. Studies addressing the increasingly pervasive goal-driven agents and robots are sparse at this point in time. This paper reviews approaches on explainable goal-driven intelligent agents and robots, focusing on techniques for explaining and communicating agents’ perceptual functions (e.g., senses, vision) and cognitive reasoning (e.g., beliefs, desires, intentions, plans, and goals) with humans in the loop. The review highlights key strategies that emphasize transparency, understandability, and continual learning for explainability. Finally, the paper presents requirements for explainability and suggests a road map for the possible realization of effective goal-driven explainable agents and robots.

The assist-as-needed technique in robotic rehabilitation is a popular technique that encourages patients’ active participation to promote motor recovery. It has been proven beneficial for patients with functional motor disability. However, its application in robotic therapy has been hindered by a poor estimation method of subjects’ functional or movement ability which is required for setting the appropriate robotic assistance. Moreover, there is also the need for consistency and repeatability of the functional ability estimation in line with the clinical procedure to facilitate a wider clinical adoption. In this study, we propose a new technique of estimation of subject’s functional ability based on the Wolf Motor Function Test. We called this estimation the functional ability index. The functional ability index enables the modulation of robotic assistance and gives a more consistent indication of subjects’ upper-limb movement ability during therapy session. Our baseline controller is an adaptive inertia-related controller, which is integrated with the functional ability index algorithm to provide movement assistance as when needed. Experimental studies are conducted on three hemiplegic patients with different levels of upper-limb impairments. Each patient is requested to perform reaching task of lifting a can from table-to-mouth according to the guidelines stipulated in the Wolf Motor Function Test. Data were collected using two inertial measurement unit sensors installed at the flexion/extension joints, and the functional ability score of each patient was rated by an experienced therapist. Results showed that the proposed functional ability index algorithm can estimate patients’ functional ability level consistently with clinical procedure and can modify generated robotic assistance in accordance with patients’ functional movement ability.

Prolong walking is a notable risk factor for work-related lower-limb disorders (WRLLD) in industries such as agriculture, construction, service profession, healthcare and retail works. It is one of the common causes of lower limb fatigue or muscular exhaustion leading to poor balance and fall. Exoskeleton technology is seen as a modern strategy to assist worker’s in these professions to minimize or eliminate the risk of WRLLDs. Exoskeleton has potentials to benefit workers in prolong walking (amongst others) by augmenting their strength, increasing their endurance, and minimizing high muscular activation, resulting in overall work efficiency and productivity. Controlling exoskeleton to achieve this purpose for able-bodied personnel without impeding their natural movement is, however, challenging. In this study, we propose a control strategy that integrates a Dual Unscented Kalman Filter (DUKF) for trajectory generation/prediction of the spatio-temporal features of human walking (i.e. joint position, and velocity, and acceleration) and an impedance cum supervisory controller to enable the exoskeleton to follow this trajectory to synchronize with the human walking. Experiment is conducted with four subjects carrying a load and walking at their normal speed- a typical scenario in industries. EMG signals taken at two muscles: Right Vastus Intermedius (on the thigh) and Right Gastrocnemius (on the calf) indicated reduction in muscular activation during the experiment. The results also show the ability of the control system to predict spatio-temporal features of the pilots’ walking and to enable the exoskeleton to move in concert with the pilot.

Assist-as-needed (AAN) robotic-rehabilitation therapy is an active area of research which aims to promote neuroplasticity and motor coordination through active participation in functional task. A key component of this strategy is to provide robotic assistance to patients only when needed. To achieve this, accurate estimation of patients’ movement/functional ability (FA) is required to evaluate patients’ need for robotic assistance and to provide the required amount of assistance, which is still a significant challenge to AAN robotic-rehabilitation therapy. This study proposes an AAN technique based on a new Functional Activity Spline Function (FASF) to estimate patients’ FA and to adapt robotic assistance. The FASF is formulated using z-spline curve to estimate patients’ movement ability based on the quality-of-movement and the time score of the patient in each functional task. A Linear Quadratic Gaussian Integral (LQGi) torque controller is applied with a FASF-to-torque mapping algorithm to physically provide low-level torque assistance on the elbow/shoulder joints. Fifteen patients were involved in the experimental study which consists of two tasks: (Task1) a pick-and-place task and (Task2) a table-to-mouth reaching task. The results showed that the proposed ANN control strategy has successfully estimated the patients’ FA consistently with high repeatability, and able to provide the robotic assistance according to the patients’ needs in the task. For different levels of impairment, the average percent-torque assistance across trials relative to the highest possible assistive torque are within the range of 5.43%–24.85% (for the mildly impaired) and 75.14%–97.14% (for the severely impaired) patents in both reaching task consistent with their FA estimation.

Human arm mechanical impedance plays a vital role in describing motion ability of the upper limb. One of the impedance parameters is stiffness which is defined as the ratio of an applied force to the measured deformation of the muscle. The arm mechanical impedance modeling is useful in order to develop a better controller for system that interacts with human as such an automated robot-assisted platform for automated rehabilitation training. The aim of the survey is to summarize the existing mechanical impedance models of human upper limb so to justify the need to have an improved version of the arm model in order to facilitate the development of better controller of such systems with ever increase in complexity. In particular, the paper will address the following issue: Human motor control and motor learning, constant and variable impedance models, methods for measuring mechanical impedance and mechanical impedance modeling techniques.