As the global population ages, there is increasing need for accessible technologies that promote cognitive health and detect early signs of cognitive decline. This research demonstrates the potential for in-residence monitoring and assessment of cognitive health using large language model (LLM)-powered socially assistive robots (SARs). We conducted a 5-week within-subjects study involving 22 older adults in retirement homes to investigate the feasibility of large language model (LLM)-powered socially assistive robots (SARs) for promoting and assessing cognitive health. We designed tasks that involved verbal dialogue based on clinically validated cognitive tools. Our findings reveal improved task performance after three robot-administered sessions, with significantly more detailed picture descriptions, fewer word repetitions in semantic fluency, and reduced need for hints. We found that older adults were more socially engaged in robot-administered tasks compared to those administered by a human, and they accepted and were willing to engage with socially assistive robots (SARs) in this context, which had not been tested before.

Prediction of transition between locomotion modes (e.g. moving from flat ground to stairs, etc) is vital for optimal interface with lower limb assistive technologies such as exoskeletons and prostheses. Inertial and bipolar electromyography (EMG) sensors have been investigated, but accuracy for clinical utility remains unresolved. This shortfall may be attributed to their limited capacity to detect subtle changes in muscle activations, particularly during the early stages of locomotion transitions (e.g., near the toe-off). In this study, we examined the effectiveness of two high-density surface electromyography (HDsEMG) sensors in detecting muscle activation changes during stair-related transitions. The results revealed that compared to bipolar EMG on the same muscles, HDsEMG-based methods increased transition prediction accuracy significantly from 70.2% to 91.1% when predicting at toe-off and from 89.8% to 99.2% when predicting with a delay of 400-ms relative to toe-off. This demonstrated the superior ability of HDsEMG to capture subtle muscle activation changes, especially during early transition stages. We also found reducing the electrode count to 21 per muscle only minimally impacted performance (88.3% accuracy at toe-off). This suggests distributing the same total number of electrodes across more muscles could potentially further improve prediction accuracy without increasing computational load. Moreover, by implementing image-inpainting signal processing, HDsEMG demonstrated robustness against the common issue of electrode signal loss. Even with 30% electrode detachment, prediction accuracy decreased only by 3%. We argue that HDsEMG offers a promising solution to bridge the gap in locomotion transition prediction for interface with assistive technology.

Functional Electrical Stimulation (FES), a key therapy for improving extremity function (e.g., in post-stroke patients), is limited by rapid FES-induced muscle fatigue. Additionally, Electromyography (EMG) monitoring is significantly compromised by FES artifacts. Mechanomyography (MMG), directly immune to such electrical FES artifacts, offers a promising alternative for fatigue estimation; however, its quantitative use for closed-loop FES remains underdeveloped. This study validated an MMG-based FES fatigue assessment system, introducing a novel wearable sensor (pressure P_MMG, microphone M_MMG) and an MMG-driven Tibialis Anterior (TA) musculotendon model with an MMG-derived fatigue index. An isometric FES fatigue protocol was conducted on control ($N=15$) and post-stroke ($N=3$) participants, recording force and MMG signals. P_MMG Mean Value (MV) signals consistently decreased with fatigue, showing strong average Pearson correlations ($\bar{r}$) with force decline in both control ($\bar{r}=0.740$) and stroke ($\bar{r}=0.928$) groups ($p \leq 0.005$). Conversely, M_MMG signals exhibited inconsistent trends and weaker force correlations, largely due to non-monotonic behavior in many participants. The P_MMG MV-driven model accurately predicted force decline, achieving mean coefficients of determination ($R^{2}$) of 0.741 (control) and 0.774 (stroke), with strong prediction correlations ($\bar{r} > 0.87, p < 0.01$). Model predictions utilizing M_MMG signals were successful only for participant subsets with consistent signal trends. The pressure-based P_MMG sensor provided a robust, non-invasive FES-induced fatigue indicator. The P_MMG-driven model allows continuous estimation of force capacity decline, promising for closed-loop FES to optimize rehabilitation.

In hand gesture recognition, classifying gestures across multiple arm postures is challenging due to the dynamic nature of muscle fibers and the need to capture muscle activity through electrical connections with the skin. This paper presents a gesture recognition architecture addressing the arm posture challenges using an unsupervised domain adaptation technique and a wearable mechanomyogram (MMG) device that does not require electrical contact with the skin. To deal with the transient characteristics of muscle activities caused by changing arm posture, Continuous Wavelet Transform (CWT) combined with Domain-Adversarial Convolutional Neural Networks (DACNN) were used to extract MMG features and classify hand gestures. DACNN was compared with supervised trained classifiers and shown to achieve consistent improvement in classification accuracies over multiple arm postures. With less than 5 minutes of setup time to record 20 examples per gesture in each arm posture, the developed method achieved an average prediction accuracy of $87.43 \%$ for classifying 5 hand gestures in the same arm posture and $64.29 \%$ across 10 different arm postures. When further expanding the MMG segmentation window from $200 \,\mathrm{ms}$ to $600 \,\mathrm{ms}$ to extract greater discriminatory information at the expense of longer response time, the intra-posture and inter-posture accuracies increased to $92.32 \%$ and $71.75 \%$. The findings demonstrate the capability of the proposed method to improve generalization throughout dynamic changes caused by arm postures during non-laboratory usages and the potential of MMG to be an alternative sensor with comparable performance to the widely used electromyogram (EMG) gesture recognition systems.

Rehabilitative robotics, particularly lower-limb exoskeletons (LLEs), have gained increasing importance in aiding patients regain ambulatory functions. One of the challenges in making these systems effective is the implementation of an assist-as-needed (AAN) control strategy that intervenes only when the patient deviates from the correct movement pattern. Equally crucial is the need for the LLE to exhibit "transparency" — minimising its interaction forces with the wearer to feel as natural as possible. This paper introduces a novel approach to transparency control based on a human-in-the-loop velocity optimisation framework. The proposed method employs torque data captured from past steps through a Series Elastic Actuator (SEA) to approximate the wearer’s intended future movements and computes a corresponding transparent velocity trajectory. The velocity commands are complemented by an Adaptive Frequency Oscillator (AFO) based position controller that leverages the periodic nature of human gait and is modified with a force sensor for increased reactiveness to human gait variations. This approach is experimentally evaluated against a standard zero-torque controller with a stationary single-degree-of-freedom knee exoskeleton test platform in a proof-of-concept study. Preliminary results indicate that combining adaptive oscillators with interaction force sensing can improve transparency compared to the conventional zero-torque controller, using force readings for position control and torque measurements for velocity optimisation and control.

To control wearable robotic systems, it is critical to obtain a prediction of the user's motion intent with high accuracy. Surface electromyography (sEMG) recordings have often been used as inputs for these devices, however bipolar sEMG electrodes are highly sensitive to their location. Positional shifts of electrodes after training gait prediction models can therefore result in severe performance degradation. This study uses high-density sEMG (HD-sEMG) electrodes to simulate various bipolar electrode signals from four leg muscles during steady-state walking. The bipolar signals were ranked based on the consistency of the corresponding sEMG envelope's activity and timing across gait cycles. The locations were then compared by evaluating the performance of an offline temporal convolutional network (TCN) that mapped sEMG signals to knee angles. The results showed that electrode locations with consistent sEMG envelopes resulted in greater prediction accuracy compared to hand-aligned placements (p$< $0.01). However, performance gains through this process were limited, and did not resolve the position shift issue. Instead of training a model for a single location, we showed that randomly sampling bipolar combinations across the HD-sEMG grid during training mitigated this effect. Models trained with this method generalized over all positions, and achieved 70% less prediction error than location specific models over the entire area of the grid. Therefore, the use of HD-sEMG grids to build training datasets could enable the development of models robust to spatial variations, and reduce the impact of muscle-specific electrode placement on accuracy.

It is well-known that the tendon-driven robotic manipulator plays an important role in robotic-assisted minimally invasive surgery (MIS). However, due to the intrinsic nonlinearities, uncertainties, slack and hysteresis introduced by the tendon-driven actuation, the tendon-driven robotic manipulator is difficult to model and control when compared with the traditional actuation styles. To serve the modeling purpose, in this paper, the deep-learning-based intelligent modeling of inverse kinematics in the snake-like tendon-driven surgical instrument is presented. In the proposed approach the Deep Recurrent Neural Network (DRNN) with Long Short-Term Memory (LSTM) architecture is adopted to memorize and identify the nonlinear inverse kinematics of the tendon-driven surgical instrument through the history of the motor and tip positions. To collect highly reliable data to train the DRNN, the experiment to generate training data is carefully designed with the consideration of the stainless tendon characters and motor limitations. During the designed controller movements, the kinematics data is obtained by recording the motor positions and the tip positions. Besides, it is noticed that there are correlations of the sequential data samples, which could significantly reduce the modeling accuracy. To remove the correlations and improve the modeling performance, the correlations of the sequential data samples are removed by modifying the training processes. Modeling results and detailed discussions verified the effectiveness of the proposed approach.

Speech/voice disorders are common in People Living with Dementia (PLwD). Fluctuations in speech quality can serve as biomarkers of cognitive deterioration but there is a gap in automated assessment of speech collected in unstructured environs. Our organisation has deployed Alexa in the households of 14 PLwD to track self-reported mental and physical state as well as use of language. n this work, we present a case study analysing highly variable speech over time, providing potential insights into cognitive changes. Alexa data gathered from the participant was manually annotated with speech assessment labels. Those labels are matched to openSMILE features by performing a feature importance analysis to isolate critical features that contribute to the perceptual ratings. We can assess phonation with a F1-score of 0.55, breathiness: 0.71, roughness: 0.60, asthenia: 0.65, strain: 0.74. This work is a first step towards automatic speech assessment to monitor cognitive impairment over time.

Conversational artificial intelligence (AI) technologies hold significant promise to support the independence, well-being and safety of older adults living with frailty or dementia at home. However, further studies are needed to identify: 1) valuable scenarios of support, 2) desired interactive features, and 3) key challenges preventing long-term adoption and utility in dementia care. In this paper, we explore the role of conversational technology in ageing and dementia care at home. Using a community-based participatory approach, we engaged 20 stakeholders, including people with lived experience of dementia and frailty, to understand preferences, perceived benefits and concerns about integrating conversational AI into daily routines at home. We uncovered key roles of the technology, including support of daily functions, health monitoring, risk mitigation, and cognitive stimulation. We emphasize the need for adapting interactions to different levels of user familiarity and progression of cognitive decline. We address the importance of the communication style and suggest careful use of open-ended questions with target populations. We further discuss feasibility considerations to overcome current barriers to adoption. Overall, this work proposes design guidelines to shape the future conceptualization and development of natural language interactions to support dementia care at home

Advancements in conversational AI have created unparalleled opportunities to promote the independence and well-being of older adults, including people living with dementia (PLWD). However, conversational agents have yet to demonstrate a direct impact in supporting target populations at home, particularly with long-term user benefits and clinical utility. We introduce an infrastructure fusing in-home activity data captured by Internet of Things (IoT) technologies with voice interactions using conversational technology (Amazon Alexa). We collect 3103 person-days of voice and environmental data across 14 households with PLWD to identify behavioural patterns. Interactions include an automated well-being questionnaire and 10 topics of interest, identified using topic modelling. Although a significant decrease in conversational technology usage was observed after the novelty phase across the cohort, steady state data acquisition for modelling was sustained. We analyse household activity sequences preceding or following Alexa interactions through pairwise similarity and clustering methods. Our analysis demonstrates the capability to identify individual behavioural patterns, changes in those patterns and the corresponding time periods. We further report that households with PLWD continued using Alexa following clinical events (e.g., hospitalisations), which offers a compelling opportunity for proactive health and well-being data gathering related to medical changes. Results demonstrate the promise of conversational AI in digital health monitoring for ageing and dementia support and offer a basis for tracking health and deterioration as indicated by household activity, which can inform healthcare professionals and relevant stakeholders for timely interventions. Future work will use the bespoke behavioural patterns extracted to create more personalised AI conversations.

We present the conceptual formulation, design, fabrication, control, and commercial translation of an Internet of Things (IoT)-enabled social robot as mapped through validation of human emotional response to its affective interactions. The robot design centers on a humanoid hybrid face that integrates a rigid faceplate with a digital display to simplify conveyance of complex facial movements while providing the impression of 3-D depth. We map the emotions of the robot to specific facial feature parameters, characterize recognisability of archetypical facial expressions, and introduce pupil dilation as an additional degree of freedom for emotion conveyance. Human interaction experiments demonstrate the ability to effectively convey emotion from the hybrid-robot face to humans. Conveyance is quantified by studying neurophysiological electroencephalography (EEG) response to perceived emotional information as well as through qualitative interviews. The results demonstrate core hybrid-face robotic expressions can be discriminated by humans (80%+recognition) and invoke face-sensitive neurophysiological event-related potentials, such as N170 and vertex positive potentials in EEG. The hybrid-face robot concept has been modified, implemented, and released in the commercial IoT robotic platform Miko (“My Companion”), an affective robot currently in use for human–robot interaction with children. We demonstrate that human EEG responses to Miko emotions are comparative to that of the hybrid-face robot validating design modifications implemented for large-scale distribution. Finally, interviews show above 90% expression recognition rates in our commercial robot. We conclude that simplified hybrid-face abstraction conveys emotions effectively and enhances human–robot interaction.

As an important movement of the daily living activities, sit-to-stand (STS) movement is usually a difficult task facing elderly and dependent people. In this article, a novel impedance modulation strategy of a lower-limb exoskeleton is proposed to provide appropriate power and balance assistance during STS movements while preserving the wearer’s control priority. The impedance modulation control strategy ensures adaptation of the mechanical impedance of the human–exoskeleton system toward a desired one requiring less wearer’s effect while reinforcing the wearer’s balance control ability during STS movements. A human joint torque observer is designed to estimate the joint torques developed by the wearer using joint position kinematics instead of electromyography or force sensors; a time-varying desired impedance model is proposed according to the wearer’s lower-limb motion ability. A virtual environmental force is designed for balance reinforcement control. Stability and robustness of the proposed method are theoretically analyzed. Simulations are implemented to illustrate the characteristics and performance of the proposed approach. Experiments with four healthy subjects are carried out to evaluate the effectiveness of the proposed method and show satisfactory results in terms of appropriate power assist and balance reinforcement.

Active lower limb prosthetics rely on the detection of gait mode to direct controller response. The majority of systems require feedback from the prosthetic and/or inertial measurement units (IMUs). Reliance on movement delays classification, reducing the range of patient activities and terrain traversed. Neuromuscular interfaces using electromyography (EMG) enable real-time monitoring by registering user intent, however EMG has known robustness issues out-of-clinic that have impeded its translation. Furthermore, supervised training of gait classifiers can require large subject-specific amputee data sets which are difficult to obtain. Mechanomyography (MMG) has shown less dependence on environmental conditions than EMG yet has seen limited use in this realm. In this investigation we introduce an MMG gait classifier targeting improved control of prosthetic (robotic) legs. We compare the accuracy of subject specific classifiers to those trained using subject-independent pooling. Additionally, we quantify the effect of introducing a small amount of data from individual test subjects to the training pool. Experiments were performed on 12 participants and 5 gait modes. A support vector machine (SVM) classifier achieved 65% accuracy with subject-specific data, 92% with pooled training data, and 94% with pooled plus limited user-specific data. The results show the promise of MMG gait classifiers with increased robustness and reduced subject-specific training in prosthetic control.

This paper introduces a novel human activity monitoring system combining Inertial Measurement Units (IMU) and Mechanomyographic (MMG) muscle sensing technology. While other work has recognised and implemented systems for combined muscle activity and motion recording, they have focused on muscle activity through EMG sensors, which have limited use outside controlled environments. MMG is a low frequency vibration emitted by skeletal muscle whose measurement does not require gel or adhesive electrodes, potentially offering much more efficient implementation for pervasive use. We have developed a combined IMU and MMG sensor the fusion of which provides a new dimension in human activity monitoring and bridges the gap between human dynamics and muscle activity. Results show that synergy between motion and muscle detection is not only viable, but straightforward, inexpensive, and can be packaged in a lightweight wearable package.

This paper introduces an inexpensive prosthetic hand control system designed to reduce the cognitive burden on amputees. It is designed around a vision-based object recognition system with an embedded camera that automates grasp selection and switching, and an inexpensive mechanomyography (MMG) sensor for hand opening and closing. A prototype has been developed and implemented to select between two different grasp configurations for the Bebionic V2 hand, developed by RSLSteeper. Pick and place experiments on 6 different objects in `Power' and `Pinch' grasps were used to assess feasibility on which to base full system development. Experimentation demonstrated an overall accuracy of 84.4% for grasp selection between pairs of objects. The results showed that it was more difficult to classify larger objects due to their size relative to the camera resolution. The grasping task became more accurate with time, indicating learning capability when estimating the position and trajectory of the hand for correct grasp selection; however further experimentation is required to form a conclusion. The limitation of this involves the use of unnatural reaching trajectories for correct grasp selection. The success in basic experimentation provides the proof of concept required for further system development.

The majority of neuroprosthetic interfaces, linking amputee to prosthetic hand, utilise proportional-based control through electromyography (EMG). The clinical translation of these interfaces can be attributed to their relative simplicity, usually requiring only two EMG electrodes to be placed on the flexor and extensor of the forearm. This bi-electrode setup enables opening and closing of hand grasp with an additional manual input used to cycle through the various grip patterns. In recent literature, the main focus has been on higher degree-of-freedom control leading to more complicated interfaces which can be considered the main barrier preventing their clinical utility. As such, new methods for grip pattern switching have not been explored with this fieldable strategy lacking any serious attention. In this work, a novel input, augmenting neuroprosthetic hand control, is proposed. This interface is based on bioacoustic signals generated through prescribed tongue movements. We demonstrate that such an interface can provide comparable performance to existing proportional-based systems without requiring any additional movements of the upper extremities.

This study aimed to improve the accuracy and robustness of a real-time assistive human machine interface system by classifying between the controlled movements related tongue-movement ear pressure (TMEP) signals and the interfering signals. The controlled movement TMEP signals were collected during left, right, up, down, flicking and pushing tongue motions. The TMEP signals were processed and classified using detection, segmentation, feature extraction and classification. The segmented signals were decomposed into the time-scale domain using a wavelet packet transform. The variance of the wavelet packet coefficients and its ratio between low-to-high scales were defined as features and the intended tongue movement commands and interfering signals were classified using both a Bayesian and support vector machine (SVM) classifiers for comparison. The average classification accuracy for discriminating between the controlled movements and the interfering signals achieved 97.8% (Bayesian) and 98.5% (SVM). The classifiers were robust remaining at a similar performance level when generalised interferences from all subjects were used. It was shown that the Bayesian classifier performed better than the SVM in a real-time environment. The approach of combining the Bayesian classifier and the wavelet packet transform provides a robust and efficient method for a real-time assistive human machine interface based on tongue-movement ear pressure signals.

This paper introduces a sensor suite framework for the partial automation of prosthetic arm control allowing high level control with a reduction of cognitive burden placed upon the user. Automation aims to replicate the hand eye co-ordination through the synergy of a virtual 7DOF arm prosthesis together with the development of a gaze tracking system. The interactions between elements of the suite are detailed and a selection of sensors implemented to control a simple simulation. Use of the novel tongue control system is used to provide discrete input to the system. Initial tests are made of of the system together with a users ability to learn to use the system with promising user feedback on ease of interaction and potential for reduced cognitive burden.

A two degree of freedom active ankle foot orthosis (AFO) for post stroke rehabilitation purposes was designed and manufactured. The AFO was cable actuated, with the force supplied by custom-built air muscles due to their low weight and compliance. A novel actuator linkage was designed in order to allow the entire actuation system to be situated behind the leg, meaning the AFO was lightweight and discrete. The novel two degree of freedom joint allowed a significantly more natural gait and comfortable user experience than conventional one degree of freedom AFOs. The effect of wearing the AFO in a passive state was analysed using a low cost inertial measurement unit (IMU), which showed that it had minimal impact on key gait parameters such as stride length and velocity. It was also trialled in a passive condition by multiple users, including 3 stroke rehabilitation physiotherapists, and was found to be comfortable to wear and easy to be put on. A parametric design process was developed, which produces a custom design based on an individual's anthropometric data. This kind of custom design would give much better conformity, decreasing skin friction and increasing comfort. It has been concluded that a custom two-degree of freedom active AFO could be of benefit after a stroke as it allows more complete rehabilitation of the ankle joint.

This paper presents a bio-inspired design of hinge joint for mobile robots based on the human knee joint. The joint mimics the curved profiles of the femur and tibia bones and also mimics the four-bar motion of the cruciate ligaments. The bio-inspired design has the same desirable features of a natural knee joint including compactness, a moving centre of rotation, high strength, high stiffness and locking in the upright position. These characteristics are important for mobile robots where there are often tight space and mass limitations. Numerical analysis and experimental tests have shown that the new hinge joint has superior performance to a pin jointed hinge in terms of stiffness and mechanical advantage.

We report the formulation and implementation of a genetic algorithm to address multi-objective optimisation of solar gain to buildings with the goal of minimising energy consumption and hence limiting carbon emissions. Heuristic optimisation approaches hold significant promise to balance complex tradeoffs in building design; however the unique nature of each building optimization problem limits broader implementation. Parameter selection is very challenging with little or no correlation between different architectural configurations. We address this issue through 'calibration' on smaller scale problems with derivable optimal solutions. Various seeding, selection and fitness options were trialled, as well as different parameter values. The Pareto front of the global solution set was successfully reproduced for the calibration case. Varying climate produced no major change in the nature of the solution; however, building orientation forced reparameterization for an optimal solution. Future work will establish when calibration is useful, and aim to quantify the nature of the solution space.

The BERT2 social robot, a platform for the exploration of human-robot interaction, is currently being built at the Bristol Robotics Laboratory. This paper describes work on the robot's face, a hybrid face composed of a plastic faceplate and an LCD display, and our implementation of facial expressions on this versatile platform. We report the implementation of two representations of affect space, each of which map the space of potential emotions to specific facial feature parameters and the results of a series of human-robot interaction experiments to characterize the recognizability of the robot's archetypal facial expressions. The tested subjects' recognition of the implemented facial expressions for happy, surprised, and sad was robust (with nearly 100% recognition). Subjects, however, tended to confuse the expressions for disgusted and afraid with other expressions, with correct recognition rates of 21.1% and 52.6% respectively. Future work involves the addition of more realistic eye movements for stronger recognition of certain responses. These results demonstrate that a hybrid face with affect space facial expression implementations can provide emotive conveyance readily recognized by human beings.

This paper presents a numerical model of a morphing wing supporting the development of a biologically inspired vehicle capable of aerial and aquatic of locomotion. The model draws inspiration from the seabird Uria aalge, the common guillemot. It is implemented within a parametric study associated with aerial and aquatic performance, specifically aiming at minimizing energy of locomotion. The implications of varying wing geometry and kinematic parameters are investigated and presented in the form of nested performance charts. Trends within both the aquatic and aerial model are discussed highlighting the implications of parameter variation on the power requirements associated with both mediums. Conflicts of geometric parameter selection are contrasted between the aerial and aquatic model, as well as other trends that impact the design of concept vehicles with this capability. The model has been validated by implementing a heuristic optimization of its key parameters under conditions akin to those of the actual bird; optimal parameters output by the model correlate to the actual behaviour of the guillemot.

Real-time implementation of an assistive human-machine interface system based around tongue-movement ear pressure (TMEP) signals is presented, alongside results from a series of simulated control tasks. The implementation of this system into an online setting involves short-term energy calculation, detection, segmentation and subsequent signal classification, all of which had to be reformulated based on previous off-line testing. This has included the formulation of a new classification and feature extraction method. This scheme utilises the discrete cosine transform to extract the frequency features from the time domain information, a univariate Gaussian maximum likelihood classifier and a two phase cross-validation procedure for feature selection and extraction. The performance of this classifier is presented alongside a real-time implementation of the decision fusion classification algorithm, with each achieving 96.28% and 93.12% respectively. The system testing takes into consideration potential segmentation of false positive signals. A simulation mapping commands to a planar wheelchair demonstrates the capacity of the system for assistive robotic control. These are the first real-time results published for a tongue-based human-machine interface that does not require a transducer to be placed within the vicinity of the oral cavity.

A strategy is introduced to rank and select principal component transform (PCT) and discrete cosine transform (DCT) transform coefficient features to overcomethe curse of dimensionalityfrequently encountered in implementing multivariate signal classifiers due to small sample sizes. The criteria considered for ranking include the magnitude, variance, interclass separation, and classification accuracies of the individual features. The feature ranking and selection strategy is applied to overcome the dimensionality problem, which often plagues the implementation and evaluation of practical Gaussian signal classifiers. The applications of the resulting PCT- and DCT-Gaussian signal classification strategies are demonstrated by classifying single channel tongue movement ear pressure signals and multichannel event related potentials. Through these experiments, it is shown that the dimension of the feature space can be decreased quite significantly by means of the feature ranking and selection strategy. The ranking strategy not only facilitates overcoming the dimensionality curse for multivariate classifier implementation but also provides a means to further select, out of a rank ordered set, a smaller set of features that give the best classification accuracies. Results show that the PCT- and DCT-Gaussian classifiers yield higher classification accuracies than those reported in previous classification studies on the same signal sets. Among the combinations of the two transforms and four feature selection criteria, the PCT-Gaussian classifiers using the maximum magnitude and maximum variance selection criteria gave the best classification accuracies across the two sets of classification experiments. Most noteworthy is the fact that the multivariate Gaussian signal classifiers developed in this paper can be implemented without having to collect a prohibitively large number of training signals simply to satisfy the dimensionality conditions. Consequently, the classification strategies can be beneficial for designing personalized human-machine interface signal classifiers for individuals from whom only a limited number of training signals can reliably be collected due to severe disabilities.

Design modifications have improved the durability and performance of a previously developed hybrid vehicle capable of both aerial and terrestrial locomotion. Whereas the original vehicle could fly, land, and crawl in sequence, it suffered from limited durability, as evidenced by catastrophic failure after a small number of landings - two to four depending on the substrate. The two most common failure modes were breakage of the terrestrial locomotion drive servos and separation of components from the fuselage. Evaluation of the original vehicle also identified the need for an autopilot. This further complicated the durability problem by greatly increasing the vehicle's mass, causing larger impulses in high speed landings. The new fuselage design includes a well-defined nacelle to which the propeller motor is securely mounted. All metal DC motors replace R/C servos in the terrestrial drive system, and a slip clutch limits the torque experienced by the motor during landing. The slip clutch comprises an annulus that drives a concentric shaft through three quad profile orings. The new 350 gram vehicle has survived eight landings on different substrates with no sign of damage.

The ultimate success of a human-robot-interface system depends on how accurately user control signals are classified. This paper is aimed at developing and testing a strategy to accurately classify human-robot control signals. The primary focus is on overcoming the dimensionality problem frequently encountered in the design of Gaussian multivariate signal classifiers. The dimensionality problem is overcome by selecting, using two different ranking criteria, a small set of linear combinations of the input signal space generated by the discrete cosine transform (DCT). The application of the resulting DCT-Gaussian signal classification strategy is demonstrated by classifying tongue-movement ear-pressure (TMEP) bioacoustic signals that have been proposed for control of an assistive robotic arm. Classification results show that the DCT-Gaussian classifiers outperform classifiers described in a previous study. Most noteworthy is the fact that the Gaussian multivariate control signal classifiers developed in this paper can be designed without having to collect a prohibitively large number of training signals in order to satisfy the dimensionality conditions. Consequently, the classification strategies will be especially beneficial for designing personalized assistive interfaces for individuals from whom only a limited number of training signals can reliably be collected due to severe disabilities.

This paper presents the tuning and implementation of a computationally efficient adaptive predictive control algorithm for robotic utility. The controller addresses the need for practical, computationally efficient, robust real-time adaptive control for multivariable robotic systems. It exploits a special matrix representation to obtain substantial reductions in the computational expense relative to standard methods. We report the design, modeling, and implementation of the controller on a simple pick-and-place manipulator and on an industrial robot loading heavy shells within the magazine of a naval vessel. The proposed controller demonstrates the ability to adapt to varying actuator performance and rapidly changing sea states. Future work involves the implementation and testing of the controller during actual naval operations. We believe this work may serve as a foundation to address control issues for robots working in uncertain dynamic environments and provide a basis for design and control of shipboard robotic devices.

The goal in this paper is to overcome the dimensionality problem related to designing human-machine-interface (HMI) signal classifiers. The dimension is decreased by selecting a small set of linear combination of the input space features using the principal components transform (PCT) and the discrete cosine transform (DCT). Issues dealing with the selection of the basis vectors of the PCT and DCT for multi-class classification problems are addressed and four different class-dependant ranking criteria are introduced to select basis vectors from the transformed training vectors in the PCT and DCT domains. The application and evaluation of the resulting PCT and DCT based multivariate classification strategies are demonstrated by classifying ear-pressure signals and event related potentials. The signals in these experiments are typical of control signals used in HMI applications and are also typical of those in which the dimensionality problem occurs. Based on the evaluations and comparisons, it is concluded that the PCT and the DCT based strategies developed in this paper offer viable solutions to overcome the dimensionality problem that frequently plagues the design of practical HMI signal classifiers.

Architecting systems designed to elicit group-level behavior beyond the capability of any single agent, however, demands a labor and experimentation-intensive cycle on the part of the programmer. As part of a system to evolve swarm behaviors, we have developed a mixed radix fitness function to overcome the problems encountered with typical fitness functions when used in a multi-objective optimization problem. In this work, we show that mixed radix fitness functions can be used to encode sequential dependencies and prioritize metrics within the context of agent-based swarm behavior. To demonstrate the effectiveness of our approach, we construct a mixed radix fitness function and evolve swarm algorithms to solve a complex extension of the classic object collection problem. Further, we show the mixed radix fitness function is successful in driving evolution towards a feasible solution while avoiding local extrema.

A new communication and control concept using tongue movements is introduced to generate, detect, and classify signals that can be used in novel hands-free human-machine interface applications such as communicating with a computer and controlling devices. The signals that are caused by tongue movements are the changes in the airflow pressure that occur in the ear canal. The goal is to demonstrate that the ear pressure signals that are acquired using a microphone that is inserted into the ear canal, due to specific tongue movements, are distinct and that the signals can be detected and classified very accurately. The strategy that is developed for demonstrating the concept includes energy-based signal detection and segmentation to extract ear pressure signals due to tongue movements, signal normalization to decrease the trial-to-trial variations in the signals, and pairwise cross-correlation signal averaging to obtain accurate estimates from ensembles of pressure signals. A new decision fusion classification algorithm is formulated to assign the pressure signals to their respective tongue-movement classes. The complete strategy of signal detection and segmentation, estimation, and classification is tested on four tongue movements of eight subjects. Through extensive experiments, it is demonstrated that the ear pressure signals due to the tongue movements are distinct and that the four pressure signals can be classified with an accuracy of more than 97% averaged across the eight subjects using the decision fusion classification algorithm. Thus, it is concluded that, through the unique concept that is introduced in this paper, human-computer interfaces that use tongue movements can be designed for hands-free communication and control applications.

A complete signal processing strategy is presented to detect and precisely recognize tongue movement by monitoring changes in airflow that occur in the ear canal. Tongue movements within the human oral cavity create unique, subtle pressure signals in the ear that can be processed to produce command signals in response to that movement. The strategy developed for the human machine interface architecture includes energy-based signal detection and segmentation to extract ear pressure signals due to tongue movements, signal normalization to decrease the trial-to-trial variations in the signals, and pairwise cross-correlation signal averaging to obtain accurate estimates from ensembles of pressure signals. A new decision fusion classification algorithm is formulated to assign the pressure signals to their respective tongue-movement classes. The complete strategy of signal detection and segmentation, estimation, and classification is tested on 4 tongue movements of 4 subjects. Through extensive experiments, it is demonstrated that the ear pressure signals due to the tongue movements are distinct and that the 4 pressure signals can be classified with over 96% classification accuracies across the 4 subjects using the decision fusion classification algorithm

The current generation of micro air vehicles (MAVs) has reached a plateau in its utility, due to reliance on single-mode locomotion. A micro vehicle capable of both aerial AND terrestrial locomotion would enjoy great utility in first responder and reconnaissance applications. To this end, BioRobots, LLC, University of Florida (UF), Case Western Reserve University, and Naval Postgraduate School (NPS) have undertaken the development of MMALV, the Morphing Micro Air-Land Vehicle. MMALV combines the UF-MAV flexible wing technology with Mini-Whegs terrestrial running gear, developed at Case. Two generations of MMALV have been developed that are capable of flying, landing, walking, and regaining flight by walking off a two-story building. To our knowledge, MMALV is the only vehicle of its size that integrates all of these functions.

A sensor platform has been developed that is capable of both aerial and terrestrial locomotion, as well as transitioning between the two. The morphing micro air-land vehicle (MMALV) implements biological inspiration in both flying and walking. MMALV integrates the University of Florida's micro air vehicle (MAV) technology with the terrain mobility of Mini-Whegs/spl trade/. Fabricated of lightweight carbon fiber, the UF-MAV employs a flexible wing design to achieve improved stability over other MAVs of similar size. Mini-Whegs/spl trade/ employs the patented (pending) wheel leg running gear that makes the Whegs/spl trade/ and Mini-Whegs/spl trade/ line of robots fast, agile, and efficient. MMALV has a 30.5cm wingspan, and is 25.4cm long. Terrestrial locomotion is achieved using two independently controlled wheel legs, which are differentially actuated to perform turning. The vehicle successfully performs the transition from flight to walking. Furthermore, MMALV is capable of transitioning from terrestrial to aerial locomotion by walking off a structure of only 20 feet. A wing retraction mechanism improves the portability of the vehicle, as well as its terrestrial stealth and ability to enter small openings.

A complete signal processing strategy is presented to detect and recognize tongue movement precisely by monitoring changes in air flow that occur in the ear canal. Tongue movements within the human oral cavity create unique, subtle pressure signals in the ear that can be processed to produce command signals in response to that movement. Once recognized, the movements can be used in human-machine interface applications, such as communicating with a computer and controlling mechanical devices. The processing strategy includes pressure signal acquisition using a microphone inserted into the ear-canal, PSD analysis to design bandpass filters to reject pressure changes due to sources other than tongue movements, start- and end-point detection in the waveforms through cross-correlation, signal estimation, and the design and evaluation of parametric and non-parametric signal classifiers. The non-parametric signal classifiers include non-linear alignment classifiers and matched filters, while the parametric classification involves a multivariate Gaussian classifier using AR model parameters. The complete strategy was tested on 4 tongue actions touching areas of the mouth: left corner; right corner; top center; bottom center. Experiments show that the pressure signals due to tongue movements are distinct and can be detected with over 97% accuracy. The unique strategy makes hands-free control of devices using tongue movements a practical reality.

A new human-machine interface is introduced for "hands-free" tele-operation of mobile robots. This interface consists of tracking tongue movement by monitoring changes in airflow that occur in the ear canal. Tongue movements within the human oral cavity create unique, subtle pressure signals in the ear that can be processed to produce commands signals in response to that movement. Once recognized, said movements can in turn be used in for robotic tele-operation. The complete strategy is tested on 4 tongue actions: touching the tongue to the left and right corners of the mouth, and to the top and bottom center of the mouth. Through extensive experiments, it is shown that the pressure signals due to tongue movements are distinct and can be detected with over 97% accuracy. A case study to control the Whegs II robotic platform has specifically been investigated. Based on simulation results, it is concluded that this unique strategy will make hands-free robotic tele-operation a practical reality.

Signal integrity closure is one of the key challenges in deep submicron physical design. In this paper, we propose a physical design methodology which includes signal integrity management through crosstalk noise analysis and repair at multiple phases of the design so that a quick noise convergence can be achieved. The methodology addresses both functional and delay noise problems in the design and is targeted for block-, platform-, and chip-level physical design of system-on-chip designs. A number of case studies are presented to illustrate the effectiveness of the proposed methodology and to provide valuable insights useful for successful signal integrity management.

A collection of agents, faced with multiple tasks to perform, must effectively map agents to tasks in order to perform the tasks quickly with limited wasted resources. We propose a decentralized control algorithm based on synchronized random number generators to enact a cooperative task auction among the agents. The algorithm finds probabilistically reasonable solutions in few rounds of bidding. Additionally, as the length of the auction increases, the expectation of a better solution increases. This algorithm is not intended to find the optimal solution; it finds a good solution with less computation and communication.

Signal integrity closure is one of the key challenges in DSM (Deep- SubMicron) physical design. In this paper, we propose a physical design methodology which includes signal integrity management through noise analysis and repair at multiple phases of the design so that a quick noise convergence can be achieved. The methodology addresses both functional and delay noise problems in the design and is targeted for block, platform, and chip level physical design of SoC (System-On-Chip) designs. A number of case studies are presented to illustrate the effectiveness of the proposed methodology and to provide valuable insights useful for successful signal integrity management.

Traditional control approaches fall well short of the necessary flexibility and efficiency needed to meet the commercial and military demands placed upon UAV swarms. Effective coordination of these swarms requires development of control strategies based on emergent behavior. We have developed a rule-based, decentralized control algorithm that relies on constrained randomized behavior and respects UAV restrictions on sensors, computation, and flight envelope. To demonstrate and evaluate the effectiveness of our approach, we have created a simulation of an air vehicle swarm searching for and mapping a chemical cloud within a patrolled region. We then consider several different detection and mapping strategies based on emergent behavior. We then establish an inverse linear relation between the size of the swarm and the time to detect the cloud, regardless of the size of the cloud. Further, we also show the size of the swarm has a linear relation with the successful detection of the cloud.

We introduce an evolutionary flight path planning algorithm capable of mapping paths for free-flying vehicles functioning under several aerodynamic constraints. An air-to-ground targeting scenario was selected to demonstrate the algorithm. The task of the path planner was to generate inputs flying a munition to a point where it could fire a projectile to eliminate a ground target. Vehicle flight constraints, path destination, and final orientation were optimized through fitness evaluation and iterative improvement of generations of candidate flight paths. Evolutionary operators comprised of one crossover operation and six mutation operators. Several cases for air-to-ground vehicle targeting have been successfully executed by the evolutionary flight path planning algorithm under challenging initial conditions. The results demonstrate that evolutionary optimization can achieve flight objectives for air vehicles without violating limits of the aircraft.

A target-seeking system for autonomous munitions in the endgame stage of flight is developed based upon a neural network model of the cockroach escape reflex. Despite significant differences in objectives, certain aspects of the cockroach escape response are consistent with desired characteristics of a target seeking system. An evolutionary target-seeking algorithm was generated to gather data to train the neural net target-seeking system. Targeting data was generated through intensive offline computing, which the target-seeking reflex was trained to reproduce on-line instantly. A linear quadratic regulator (LQR) autopilot executes reflexive guidance commands. With the trained target-seeking system installed on a candidate air to-ground munition, simulations show that the reflex may react to strike targets very quickly. Context dependency was demonstrated through the actions of the reflex striking targets moving on rapidly changing, evading, and unpredictable trajectories, as well as through false and disruptive sensor data.

We present a methodology for the design and analysis of power grids in the PowerPC™ microprocessors. The methodology covers the need for power grid analysis across all stages of the design process. A case study showing the application of this methodology to the PowerPC™ 750 microprocessor is presented.