Despite the global mental health crisis, access to screenings, professionals, and treatments remains high. In collaboration with licensed psychotherapists, we propose a C onversational AI T herapist with psychotherapeutic I nterventions (CaiTI), a platform that leverages large language models (LLMs) and smart devices to enable better mental health self-care. CaiTI can screen the day-to-day functioning using natural and psychotherapeutic conversations. CaiTI leverages reinforcement learning to provide personalized conversation flow. CaiTI can accurately understand and interpret user responses. When the user needs further attention during the conversation, CaiTI can provide conversational psychotherapeutic interventions, including cognitive behavioral therapy and motivational interviewing. Leveraging the datasets prepared by the licensed psychotherapists, we experiment and microbenchmark various LLMs’ performance in tasks along CaiTI’s conversation flow and discuss their strengths and weaknesses. With the psychotherapists, we implement CaiTI and conduct 14-day and 24-week studies. The study results, validated by therapists, demonstrate that CaiTI can converse with users naturally, accurately understand and interpret user responses, and provide psychotherapeutic interventions appropriately and effectively. We showcase the potential of CaiTI LLMs to assist the mental therapy diagnosis and treatment and improve day-to-day functioning screening and precautionary psychotherapeutic intervention systems.

Cardio exercise elevates both heart rate and respiration rate, resulting in distinct physiological changes that affect speech patterns, pitch, breathing sounds, and heart sounds. These variations, which occur post-exercise, are influenced by factors such as exercise intensity and individual fitness levels. A comprehensive audio dataset is critically needed to capture post-exercise physiological changes, as existing datasets focus mainly on resting speech, breathing, and heart sounds, neglecting the dynamic shifts following physical exertion. Current datasets fail to capture unique post-exercise variations like speech disfluencies, altered breathing patterns, and variable heart sound intensities, limiting model generalizability to post-exercise conditions. To address this gap, we recruited 59 subjects from diverse backgrounds to engage in cardio exercise, specifically running, reaching varied exertion levels to produce a rich dataset. Our dataset includes 250 sessions totaling 143 minutes of structured reading, 47 minutes of spontaneous speech, 71 minutes of breathing sounds, and 62.5 minutes of phonocardiogram (PCG) recordings. We designed and deployed preliminary case studies to show that speech changes post-cardio could serve as an indicator of exertion level. We envision this dataset as a foundational resource for designing models in speech and cardiorespiratory monitoring that are resilient to the physiological shifts induced by exercise. This dataset could advance natural language processing (NLP) applications, mobile health, and wearable sensing technologies by enabling resilient and accurate physiological monitoring in real-world conditions.

We investigate the conducted electromagnetic interference (EMI) emissions for a variable-frequency, soft-switching, half-bridge dc-ac inverter. We propose a method to determine the maximum EMI emission setpoint, then determine the greatest required attenuation to comply with CISPR25 2021 Class 5 conducted emissions limits in a simulated electromagnetic precompliance testing process. By designing a single filter for the worst dc-dc emission setpoint, we hypothesize that it will satisfy all of the other setpoints during variable frequency dc-ac operation. The circuit is simulated in LTSpice at 81 setpoints, and EMI measurements are taken with a simulated CISPR25-compliant line impedance stabilization network (LISN). Heatmaps of the operation setpoints display maximum required EMI attenuation, filter cutoff frequency, and filter LC product. Then, EMI measurements are plotted and compared to the limits before and after a filter is applied. By designing an EMI filter for the single greatest emission setpoint of a variable-frequency, soft-switching dc-ac inverter, all other operating points also pass precompliance testing.

Minimization of electrical losses throughout full machine operation is essential for applications of modern drives with wide operating ranges such as vehicle drivetrains. Producing optimal reference currents that minimize copper loss and core loss for a combination of torque and speed is generally a difficult problem to solve analytically given the many non-linearities in a machine. The wound rotor synchronous machine’s magnetic behavior changes between zero torque and rated torque. The machine considered in this study has a saliency (Ld> Lq) at zero torque and a saliency (Ldq) at rated torque. An affine magnetics model is created at each of these two points. Using the zero torque inductance approximation to predict the flux at the peak torque operating point leads to an error of 278 μVs or 94% for d-axis flux and 101 μVs or 80% for q-axis flux. Additionally, use of a constrained nonlinear optimization solver results in different sets of optimal efficiency reference currents for the two models.

The dual active half-bridge (DAHB) converter allows multiple power sources to be interfaced together while maintaining simple control and high efficiency. This paper presents a novel analysis of the DAHB, with a focus on analytical modeling and dc current control. Departing from prior research, a steady-state model is proposed to compute the current wave-forms in the transformer and across the four ports of the converter using a superposition of effects approach. Next, a dc current controller is developed to manage the transformer currents and to achieve the desired currents in the four capacitors. The effectiveness of the proposed model and control algorithm is tested using both simulations and experimental data.

Over half a trillion recreational runners worldwide engage in running for psychological, health, and social benefits. Running metrics are essential for motivation, goal setting, performance improvement, health management, and injury prevention. Although wearable devices like fitness trackers and smartwatches offer various metrics, they often perform poorly on treadmills and can be uncomfortable or restrictive. In this work, we propose a non-contact, real-time, smartphone-based approach to estimate running metrics, including cadence, ground contact time (GCT), and balance, using the sound produced during treadmill running. In collaboration with a licensed running coach, we recruited over 50 subjects with varying levels of running expertise. We collected treadmill running sounds and ground-truth running metrics in different environments. We designed and developed a multi-task learning (MTL) machine learning mobile system to capture the treadmill running sounds and estimate running metrics in situ. Our proposed method shows comparable accuracy in estimating running metrics to commercial off-the-shelf (COTS) wearable devices.

Traditionally, Dual Active Half Bridge (DAHB) converters are operated in a restricted number of modes that focus on power transfer across the central transformer. This research identifies all possible modes of the DAHB converter’s operation and how the modes affect what power is transferred between specific capacitors. Analysis begins by deconstructing the DAHB into two half bridge circuits that dictate power transfer between upper and lower capacitors and a DAHB that is restricted to power transfer across the transformer. The operating modes of each of the circuits are examined by considering the permutations of the switching instances, which are related to the phase shifts and duty cycles of each pair of switches, in every period. These permutations reveal the operating modes. Models for the HB and DAHB are created; all modes of operation are captured and the models are verified against Matlab-PLECS simulations.

This paper proposes a multilevel power converter topology that processes less power than it outputs. It can be scaled to N-levels with linear component quantity and stress scaling and voltage balance is maintained for any voltage conversion ratio. It is composed of a set of series stacked capacitors where each additional capacitor defines an additional voltage level. Input voltage is applied across the entirety of the capacitor stack and the output can be taken at any node between capacitors. Capacitor voltage balance is maintained through any method of energy sharing between capacitors. The amount of power that needs to be transferred between capacitors to maintain voltage balance is less than the output power of the converter. Equations that describe this required power transfer are derived. An example implementation of a capacitive power transfer mechanism is shown. Functionality of the proposed topological framework is proven through high-fidelity simulation of an 8-level converter consisting of 8 series capacitors in both DC/DC and DC/AC modes of operation.

In this demonstration, in collaboration with licensed therapists, we introduce an AI therapist that takes advantage of the smart-home environment to screen day-to-day functioning and infer mental wellness of an occupant. Our system can assess a user's daily functioning and mental wellness based on a combination of direct conversation with users and information obtained from smart home devices using psychological rubrics proposed in [1]. We demonstrate that our system can converse with a user in a natural way (through a smartphone or smart speaker) and analyze a user's response semantically and sentimentally. In addition, we show that our system can provide preliminary interventions to help improve the user's wellness. In particular, when abnormal behavior is detected during the conversation or by smart home devices, the system provides psychotherapeutic consolations during the conversation and will check on the occupant's condition by actuating a home robot.

This paper proposes a novel multi-level power converter topology that is an adaptation of existing topologies. This topology is fully balanced, can function bidirectionally for both DC/AC and DC/DC modes of operation, and can be expanded to an unrestricted N levels. It is comprised of discrete switching cells across which the input voltage can be arbitrarily distributed. Each switching cell can operate as a standalone unit and is individually controllable. Fundamental equations for each cell and the converter as a whole are derived alongside methods for expanding to N levels. Validation of this topology and its fundamental equations are shown through high-fidelity simulation of a 7-level converter that is comprised of five individual switching cells. Two example methods of determining duty cycle combinations are explored, one basic and one optimized, and the characteristics are discussed. Converter performance as N approaches ∞ are provided.

In this paper, a virtual-flux finite control set model predictive control (FCS-MPC) strategy of switched reluctance motor (SRM) drives is proposed. This technique uses a flux linkage-tracking algorithm to indirectly control the phase current. The algorithm is based on an estimated virtual flux obtained from the static characteristics of the machine. A cost function is used to evaluate the switching state that produces the minimum error. A state graph for switching states limitation is also proposed to reduce the number of commutations and computational burden. Simulation results evidence the enhanced performance of the proposed technique with respect to hysteresis control for current tracking using two different current shaping techniques: torque sharing function (TSF) and radial force shaping (RFS).

Moving Horizon Estimators (MHE)solve an optimization problem to estimate unknown states or parameters based on a sequence of measurements containing disturbances and noise in nonlinear systems subject to input and state constraints. This research applies MHE to estimating the nonlinear behavior of interior-mount permanent magnet synchronous machines (IPMSM). The MHE estimates a disturbance term that reppresents the speed-dependent nonlinear terms of the machine model and can be interpreted as extended back-EMF. The formulation is based on a cost function that matches the measurements to the model formulation and an explicit regularization penalty is added. The term is interpreted as a gain that balances estimation accuracy with noise rejection. We demonstrate that our formulation can be solved in realtime and is effective in estimating the disturbance term on an experimental test bench both in terms of noise rejection and estimation accuracy.

This paper proposes a simplified convex optimization-based sensorless scheme for interior permanent magnet synchronous motor (IPMSM)drives. The computational burden of the existing convex optimization-based method is significantly reduced by a single variable cost function, which is based on the machine voltage equations in the stationary reference frame. With less computation, the proposed method provides good performance characteristics similar to the existing one, e.g., dynamic speed response and smooth transition between the low and high-speed ranges. The convergence capability of the cost function is also confirmed by a convexity analysis. The feasibility of the control technique is experimentally validated in a test bench, demonstrating the accuracy of the technique and its reduced computational burden.

The active or passive decoupling method has to be utilized to deal with the second-order harmonic existing in the DC-bus of the grid-tied single-phase inverters. Compared with the active decoupling method, the passive decoupling method is simpler, cheaper and more reliable. The electrolytic capacitors are usually used in the DC-bus as typical passive decoupling components. The film capacitors can be added in parallel with the electrolytic capacitor to help filtering out the high frequency harmonics to extend the electrolytic capacitors' life. In addition, the LC resonant filter can be utilized for the decoupling purpose to achieve better performance. However due to the relatively low resonant frequency, it results in large inductance which will significantly increase the size and cost of the system. A current sharing method is proposed in this paper. With this method, an inductor with reasonable size can be utilized in the LC resonant filter to further extend the electrolytic capacitors' life. In this paper, the design procedure of the hybrid capacitor bank for the single-phase inverter with unipolar modulation will be discussed. The simulation and experimental results will be provided to verify the design of the hybrid capacitor bank for a 3kW single-phase PV inverter.

This paper presents the low speed position estimation scheme for an IPM machine controlled by model predictive control with finite control set. The career signal injection is not viable as there is no PWM to superimpose it with PWM for this type of control. The pulse vector injection technique requires current derivative sensors which makes the overall scheme less attractive. This paper utilizes the inherent high frequency vector injection of the model predictive control to extract the position information. It is shown that the high frequency current response is amplitude modulated with respect to the position. A demodulation technique based on the reactive power estimation is proposed. The simulation results at various initial positions confirm the validity of the proposed position estimation scheme.

A battery-ultracapacitor Hybrid Energy Storage System (HESS) combines the advantages of both Li-ion batteries and ultracapacitors. Li-ion batteries sustain a relatively long electric only driving range but degrade if exposed to high C-rates and large number of cycles. Ultracapacitors are robust, have a quasi infinite cycle life and can sustain highly dynamic power profiles. This paper proposes a HESS Linear Quadratic Regulator (LQR) design to mitigate issues related to battery wear and peak power demands for electric and hybrid electric vehicles. The LQR controller imposes the battery current with a bidirectional power electronic converter that interfaces the battery to the ultracapacitor. The HESS is accurately modeled using experimental battery and ultracapacitor data in conjunction with equivalent circuit models. Simulations are carried out to validate the LQR controller on a UDDS drive cycle. Reduced battery wear is quantified using a spectral analysis of the battery current, which identifies microcycles.

Online estimation of losses is important to improve control, operation and monitoring of electrical machines. This paper focuses on investigation of iron losses using a magnetic circuit model for its accuracy and adaptability to online estimation purposes. A customized magnetic circuit of an IPM machine is proposed that captures slotting, non linearity, cross saturation and localized effect of flux bridges. A technique is presented to compute the alternating stator and rotor flux from the static magnetic circuit. The required computations are reduced introducing pseudo mmf sources that avoid solving magnetic circuit at several rotational steps. The stator and rotor core losses are found corresponding to each harmonic component of alternating flux density. The results are validated with Finite Element analysis with good correlation.

Thermoelectric Generators (TEG) can harvest a part of the thermal energy otherwise lost in the exhaust gases of vehicles and are combined with Maximum Power Point Tracking (MPPT) schemes to maximize the power output. This paper proposes a novel TEG MPPT scheme named High Frequency Injection (HFI) method. The method injects a high frequency voltage to the TEG and yields a power with a high frequency component. This component is demodulated and yields a signal proportional to the distance from the optimal operation point. The duty cycle setpoint is adjusted with a proportional-integral (PI) controller. The method is compared to the Perturb & Observe method using a drive cycle. Both show good results in terms of dynamic tracking of the optimal operation point. However, the HFI method is shown to be significantly more robust against sensor noise.

This paper proposes a novel initial rotor position estimation algorithm for Interior Permanent Magnet Synchronous Machine (IPMSM) drives. First, the rotor position is determined based on the machine saliency using the flux equations in the stationary reference frame. Since the machine saliency performs two periods in one electrical cycle, there exists an ambiguity of 180° in the estimation result. The location of the magnetic north pole is detected using a generalized polarity detection method. This method injects voltage pulses and compares the current response with the expected response using the d-axis differential inductance profile. An accurate nonlinear machine model is introduced for analysis and simulation of sensorless control in IPMSM drives. The model uses the machine flux as dynamic equation and the flux current relationship as output function avoiding approximations due to saturation. The initial position detection procedure is validated with this model using experimental current-flux data.

The direct torque control approach called model predictive direct torque control (MP-DTC) is extended to field-weakening operation in this research. The controller is of the finite control set (FCS) type, which takes the discrete states of the voltage source inverter (VSI) into account. Each sampling period, the future behavior of the plant is predicted and inputs (voltage vectors) are selected using an optimization criterion, i.e., a cost function. Optimization, however, depend on the operation region. Maximum torque per ampere (MTPA) tracking for high electrical efficiency is aimed below rated speed, operation off the MTPA trajectory is necessary for obtaining field-weakening. In this work, a cost function is designed, which is suitable for operation at high speeds without penalizing operation below rated speed. MP-DTC is applied to permanent magnet synchronous machine (PMSM) drive systems. High control performances, i.e., dynamics, is obtained without increasing the switching frequency nor reducing significantly the torque quality. This feature is interesting above all for high power applications. The proposed control strategy has been evaluated on a small-scale (PMSM) drive system with two-level VSI for demonstration showing promising results.

A novel type of Model Predictive Direct Torque Control (MP-DTC) is proposed in this paper. The future behavior of the plant is predicted using a discrete-time state-space model in the dq reference frame for a finite set of voltage vectors, which is pre-selected by a graph algorithm. The preferable input is chosen based on a cost function. Contrarily to conventional MP-DTC (and DTC), the stator flux is not explicitly controlled nor hysteresis bounds are used. The cost function is designed to meet multiple demands: torque reference tracking, Maximum Torque per Ampere (MTPA) tracking for high electrical efficiency, and limitation of the states to their maximum admissible values.