Episodic tasks in Reinforcement Learning (RL) often pose challenges due to sparse reward signals and high-dimensional state spaces, which hinder efficient learning. Additionally, these tasks often feature hidden “trap states”—irreversible failures that prevent task completion but do not provide explicit negative rewards to guide agents away from repeated errors. To address these issues, we propose Time-Weighted Contrastive Reward Learning (TW-CRL), an Inverse Reinforcement Learning (IRL) framework that leverages both successful and failed demonstrations. By incorporating temporal information, TW-CRL learns a dense reward function that identifies critical states associated with success or failure. This approach not only enables agents to avoid trap states but also encourages meaningful exploration beyond simple imitation of expert trajectories. Empirical evaluations on navigation tasks and robotic manipulation benchmarks demonstrate that TW-CRL surpasses state-of-the-art methods, achieving improved efficiency and robustness.

Achieving multi-lead electrocardiography (ECG) in consumer-grade wearables remains challenging, as most devices provide only a few electrodes and cannot capture spatially diverse cardiac signals. Conventional 12-lead ECG, while clinically standard, requires ten electrodes across the body, confining its use to medical environments. We present EarSleeve, a modular dual-electrode eartip sleeve that transforms off-the-shelf earphones into a 12-lead ECG sensing platform through a human-in-the-loop design. Each sleeve embeds two conductive electrodes and electrically links both sides to form a four-electrode configuration. EarSleeve simultaneously records six limb leads and reconstructs 12-lead–equivalent ECG signals by sequentially contacting standard chest locations. In a 12-user study, EarSleeve captures clear ECG waveforms across all leads and is evaluated against an FDA-cleared reference. Results demonstrate the feasibility of reconstructing 12-lead–equivalent ECG signals using a minimum-electrode configuration under controlled conditions. To our knowledge, EarSleeve is the first system to achieve this with off-the-shelf earphones.

Deploying Large Language Models (LLMs) directly on mobile platforms at the edge is gaining a lot of traction due to a myriad of benefits, such as increased privacy, personalization, and reduced latency. However, LLMs have heavy computational requirements, which are difficult for resource-constrained mobile and edge platforms to fulfill. In addition to limited compute resources, mobile and edge systems often have a compact form factor and lack physical mechanisms to dissipate heat generated from high processor usage rates (e.g., fans) to prevent throttling and reduced processing power, which LLMs can easily cause. To mitigate these effects, prior works have proposed various power governing strategies, such as dynamic voltage and frequency scaling (DVFS), for reducing power and heat generation for heavy computational tasks on mobile platforms. Recently, DVFS methods tailored for mobile LLMs have also been proposed. However, these methods mostly focus on optimizing hardware parameters and processor frequencies, and they fall short under some thermally constrained scenarios. Drawing from recent advances in machine learning, we identify and take advantage of the key insight that not all tokens require full-depth inference to maintain high-quality generation. Motivated by this, we present PELM, a solution that augments traditional DVFS processor frequency tuning with two additional workload-specific knobs: 1) speculative decoding and 2) variable verification depth to expand the optimization space to multiple dimensions for more power efficient on-device LLM inference. In extensive evaluations across different hardware platforms and datasets, PELM demonstrates superior performance compared to state-of-the-art power governing methods, with up to 23.1% speedup and 52.4% reduction in energy consumption, while maintaining comparable task performance. The source code is available at https://github.com/imec-nu/PELM.

Large Language Models (LLMs) have shown immense human-like capabilities for reasoning and generating digital content. However, their ability to freely sense, interact, and actuate the physical domain remains significantly limited due to three fundamental challenges: (1) physical environments require specialized sensors for different tasks, yet deploying dedicated sensors for each application is impractical; (2) events and objects of interest are often localized to small areas within large spaces, making them difficult to detect with static sensor networks; and (3) foundation models need flexible actuation capabilities to meaningfully interact with the physical world. To bridge this gap, we introduce EmbodiedFly, an embodied LLM agent combining a foundation model pipeline with a reconfigurable drone platform to observe, understand, and interact with the physical world. Our co-design approach features (1) a FM orchestration framework connecting multiple LLMs, VLMs, and an open-set object detection model; (2) a novel image segmentation technique that identifies task-relevant areas; and (3) a custom drone platform that autonomously reconfigures with appropriate sensors and actuators based on commands from the FM orchestration framework. Through real-world deployments, we demonstrate that EmbodiedFly completes diverse physical tasks with up to \(85\%\) higher success rates compared to traditional approaches leveraging static deployments.

Accuracy remains a standard metric for evaluating AI systems, but it offers limited insight into how models arrive at their solutions. In this work, we introduce a benchmark based on brainteasers written in long narrative form to probe more deeply into the types of reasoning strategies that models use. Brainteasers are well-suited for this goal because they can be solved with multiple approaches, such as a few-step solution that uses a creative insight or a longer solution that uses more brute force. We investigate large language models (LLMs) across multiple layers of reasoning, focusing not only on correctness but also on the quality and creativity of their solutions. We investigate many aspects of the reasoning process: (1) semantic parsing of the brainteasers into precise mathematical competition style formats; (2) self-correcting solutions based on gold solutions; (3) producing step-by-step sketches of solutions; and (4) making use of hints. We find that LLMs are in many cases able to find creative, insightful solutions to brainteasers, suggesting that they capture some of the capacities needed to solve novel problems in creative ways. Nonetheless, there also remain situations where they rely on brute force despite the availability of more efficient, creative solutions, highlighting a potential direction for improvement in the reasoning abilities of LLMs.

From clinical healthcare to daily living, continuous sensor monitoring across multiple modalities has shown great promise for real-world intelligent decision-making but also faces various challenges. In this work, we argue for modeling such heterogeneous data sources under the multimodal paradigm and introduce a new framework, MAESTRO. We introduce MAESTRO, a novel framework that overcomes key limitations of existing multimodal learning approaches: (1) reliance on a single primary modality for alignment, (2) pairwise modeling of modalities, and (3) assumption of complete modality observations. These limitations hinder the applicability of these approaches in real-world multimodal time-series settings, where primary modality priors are often unclear, the number of modalities can be large (making pairwise modeling impractical), and sensor failures often result in arbitrary missing observations. At its core, MAESTRO facilitates dynamic intra- and cross-modal interactions based on task relevance, and leverages symbolic tokenization and adaptive attention budgeting to construct long multimodal sequences, which are processed via sparse cross-modal attention. The resulting cross-modal tokens are routed through a sparse Mixture-of-Experts (MoE) mechanism, enabling black-box specialization under varying modality combinations. We evaluate MAESTRO against 10 baselines on four diverse datasets spanning three applications, and observe average relative improvements of 4% and 8% over the best existing multimodal and multivariate approaches, respectively, under complete observations. Under partial observations—with up to 40% of missing modalities—MAESTRO achieves an average 9% improvement. Further analysis also demonstrates the robustness and efficiency of MAESTRO's sparse, modality-aware design for learning from dynamic time series.

Foundation models (FM) have shown immense human-like capabilities for generating digital media. However, foundation models that can freely sense, interact, and actuate the physical domain is far from being realized. This is due to 1) requiring dense deployments of sensors to fully cover and analyze large spaces, while 2) events often being localized to small areas, making it difficult for FMs to pinpoint relevant areas of interest relevant to the current task. We propose FlexiFly, a platform that enables FMs to "zoom in" and analyze relevant areas with higher granularity to better understand the physical environment and carry out tasks. FlexiFly accomplishes by introducing 1) a novel image segmentation technique that aids in identifying relevant locations and 2) a modular and reconfigurable sensing and actuation drone platform that FMs can actuate to "zoom in" with relevant sensors and actuators. We demonstrate through real smart home deployments that FlexiFly enables FMs and LLMs to complete diverse tasks up to 85% more successfully. FlexiFly is critical step towards FMs and LLMs that can naturally interface with the physical world.

Speech super-resolution techniques offer a promising solution to enhance audio quality in wearable devices, particularly when addressing the challenges of reduced sampling rates necessitated by battery life constraints and network instability. However, existing methods either prove computationally prohibitive for mobile platforms, or lack sufficient performance. We present TraMSR, a novel hybrid model combining transformer and Mamba architectures for acoustic speech super-resolution. TraMSR achieves superior performance while significantly reducing computational demands compared to state-of-the-art methods. Our model outperforms GAN-based approaches by up to 7.3% in Perceptual Evaluation of Speech Quality (PESQ) and 1.8% in Short-Time Objective Intelligibility (STOI), with an order of magnitude smaller memory footprint and up to 465 times faster inference speed.

Foundation models excel in tasks such as content generation, zero-shot classifications, and reasoning. However, they struggle with sensing, interacting, and actuating in the physical world due to their dependence on limited sensors and actuators in providing timely contextual information or physical interactions. This reliance restricts the system's adaptability and coverage. To address these issues and create an embodied AI with foundation models (FMs), we introduce Embodied Reconfigurable Drone Agent (EmbodiedRDA). EmbodiedRDA features a custom drone platform that can autonomously swap payloads to reconfigure itself with a diverse list of sensors and actuators. We designed FM agents to instruct the drone to equip itself with appropriate physical modules, analyze sensor data, make decisions, and control the drone's actions. This enables the system to perform a variety of tasks in dynamic physical environments, bridging the gap between the digital and physical worlds.

The stethoscope is one of the most important diagnostic tools used by healthcare professionals, through a process called auscultation, to screen patients for abnormalities of the heart and lungs. While there are digital stethoscopes on the market which ease this process, it still takes years of training to properly use these devices to listen for abnormal sounds within the body. We present ARSteth, an intelligent stethoscope platform that improves the accessibility of stethoscopes for the general population, allowing anyone to perform auscultation in the comfort of their own homes. Our platform utilizes a combination of augmented reality (AR), acoustic intelligence, and human-machine interaction to dynamically guide users on where to place the stethoscope on different parts of the body (auscultation points), through visual and audio cues. Through user studies, we show that ARSteth, on average, can guide users within 13.2 mm from optimal auscultation points marked by licensed physicians in 13.09 seconds for each auscultation point. By guiding users towards more effective auscultation points, make preventative health screening more accessible and effective for everyone we are able to achieve higher confidence on classifying heart murmurs.

Mapping 3D airflow fields is important for many HVAC, industrial, medical, and home applications. However, current approaches are expensive and time-consuming. We present Anemoi, a sub-$100 drone-based system for autonomously mapping 3D airflow fields in indoor environments. Anemoi leverages the effects of airflow on motor control signals to estimate the magnitude and direction of wind at any given point in space. We introduce an exploration algorithm for selecting optimal waypoints that minimize overall airflow estimation uncertainty. We demonstrate through microbenchmarks and real deployments that Anemoi is able to estimate wind speed and direction with errors up to 0.41 m/s and 25.1° lower than the existing state of the art and map 3D airflow fields with an average RMS error of 0.73 m/s.

We present SoFIT, an easily-deployed and privacy-preserving camera network system for occupant tracking. Unlike traditional camera network-based systems, SoFIT does not require a person to calibrate the network or provide real-world references. This enables anyone, including non-professionals, to install SoFIT. Once installed, SoFIT automatically localizes cameras within the network and generates the floor map leveraging movements of people using the space in daily life, before using the floor map and camera locations to track occupants throughout the environment. We demonstrate through a series of deployments that SoFIT can localize cameras with less than 4.8cm error, generate floor maps with 85% similarity to actual floor maps, and track occupants with less than 7.8cm error.

With the recent societal impact of COVID-19, companies and government agencies alike have turned to thermal camera based skin temperature sensing technology to help screen for fever. However, the cost and deployment restrictions limit the wide use of these thermal sensing technologies. In this work, we present SIFTER, a low-cost system based on a RGB-thermal camera for continuous fever screening of multiple people. This system detects and tracks heads in the RGB and thermal domains and constructs thermal heat map models for each tracked person, and classifies people as having or not having fever. SIFTER can obtain key temperature features of heads in-situ at a distance and produce fever screening predictions in real-time, significantly improving screening through-put while minimizing disruption to normal activities. In our clinic deployment, SIFTER measurement error is within 0.4°F at 2 meters and around 0.6°F at 3.5 meters. In comparison, most infrared thermal scanners on the market costing several thousand dollars have around 1°F measurement error measured within 0.5 meters. SIFTER can achieve 100% true positive rate with 22.5% false positive rate without requiring any human interaction, greatly outperforming our baseline [1], which sees a false positive rate of 78.5%.

Audio is valuable in many mobile, embedded, and cyber-physical systems. We propose AvA, an acoustic adaptive filtering architecture, configurable to a wide range of applications and systems. By incorporating AvA into their own systems, developers can select which sounds to enhance or filter out depending on their application needs. AvA accomplishes this by using a novel adaptive beamforming algorithm called content-informed adaptive beam-forming (CIBF), that directly uses detectors and sound models that developers have created for their own applications to enhance or filter out sounds. CIBF uses a novel three step approach to prop-agate gradients from a wide range of different model types and signal feature representations to learn filter coefficients. We apply AvA to four scenarios and demonstrate that AvA enhances their respective performances by up to 11.1%. We also integrate AvA into two different mobile/embedded platforms with widely different resource constraints and target sounds/noises to show the boosts in performance and robustness these applications can see using AvA.

With the global spread of the COVID-19 pandemic, ventilation indoors is becoming increasingly important in preventing the spread of airborne viruses. However, while sensors exist to measure wind speed and airflow gradients, they must be manually held by a human or an autonomous vehicle, robot, or drone that moves around the space to build an airflow map of the environment. In this demonstration, we present DAE, a novel drone-based system that can automatically navigate and estimate air flow in a space without the need of additional sensors attached onto the drone. DAE directly utilizes the flight controller data that all drones use to self-stabilize in the air to estimate airflow. DAE estimates airflow gradients in a room based on how the flight controller adjusts the motors on the drone to compensate external perturbations and air currents, without the need for attaching additional wind or airflow sensors.

There has been an immense growth in sensors, actuators, and smart devices in recent years, which enable us to better sense, actuate, and understand the physical world. Despite this growth, we have yet to achieve fully intelligent environments. This is, in part, due to the large number of different organizations creating smart devices with proprietary technologies and communication protocols that are not compatible with each other and require significant engineering to incorporate and adapt to specific applications. In this work, we present an easy-to-install and low-cost embedded platform that allows users to rapidly configure a mixture of sensors and actuators. The system is based on the commonly-used Raspberry Pi ecosystem, easily configurable, and does not require users to have prior knowledge of programming, which allows anyone, regardless of background, to use. We also introduce a battery-powered wireless extension module that is suitable for mobile drone applications, where a chord-powered Raspberry Pi is not suitable. We demonstrate the impact our system has on enabling drones with flexible sensing modalities and creating smarter environments by integrating our platform into a variety of intelligent home applications.

Cardiopulmonary ailments are a major cause of mortality. Stethoscopes are one of the most important tools that healthcare professionals use to screen patients for a variety of ailments, especially those related to the heart and lungs. Despite the growth of digital stethoscopes on the market, it takes years of training to properly use stethoscopes to listen for abnormal sounds within the body. In this demonstration, we present an intelligent stethoscope platform that makes stethoscopes more accessible to the general population. Our platform utilizes augmented reality (AR) to provide real-time guidance on where to properly place the stethoscope on the body, enabling the general population to screen themselves for ailments.

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.

Sound detection and classification are critical in many acoustic-based applications. Existing works generally focus on discovering new features and classifiers to improve detection. However, in many scenarios the presence of other sounds may hinder the performance of these sound classifiers. In this work, we take a sound filtering and enhancement approach to improve sound detection for mobile and embedded applications, regardless of the type of detector used.

Vehicle accidents are one of the greatest cause of death and injury in urban areas for pedestrians, workers, and police alike. In this work, we present CSafe, a low power audio-wearable platform that detects, localizes, and provides alerts about oncoming vehicles to improve construction worker safety. Construction worker safety is a much more challenging problem than general urban or pedestrian safety in that the sound of construction tools can be up to orders of magnitude greater than that of vehicles, making vehicle detection and localization exceptionally difficult. To overcome these challenges, we develop a novel sound source separation algorithm, called Probabilistic Template Matching (PTM), as well as a novel noise filtering architecture to remove loud construction noises from our observed signals. We show that our architecture can improve vehicle detection by up to 12% over other state-of-art source separation algorithms. We integrate PTM and our noise filtering architecture into CSafe and show through a series of real-world experiments that CSafe can achieve up to an 82% vehicle detection rate and a 6.90° mean localization error in acoustically noisy construction site scenarios, which is 16% higher and almost 30° lower than the state-of-art audio wearable safety works.

Homes are becoming more intelligent due to the growth of smart sensors and devices found in typical homes. However, most of these sensors and devices function independently from one another, limiting the amount of utility and services a truly "smart" home can provide. In this demonstration, we introduce two key ideas towards more intelligent homes. First, we explore the usage of mobile drones in the home environment. Second, we propose DIA, a system that seamlessly connects to the home environment and automatically discovers and jointly utilizes smart sensors and actuators around the home to provide services that are otherwise not possible. We demonstrate three services that DIA enables.

Facial expression monitoring is crucial in fields including mental health care, driver assistant systems, and advertising. However, existing systems typically rely on cameras that capture entire faces, or contact-based bio-signal sensors, which are neither comfortable nor portable. In this demonstration, we present a wireless glasses system for non-contact facial expression monitoring. The system is composed of an IR camera and an embedded processing unit mounted on a 3D-printed glasses frame, and a novel data processing pipeline running across the glasses platform and a computer. Our system performs high-accuracy and real-time facial expression detection with a running time of up to 9 hours. We will show the fully-functioning wearable system in this demonstration.

In this article, we present recEnergy, a recommender system for reducing energy consumption in commercial buildings with human-in-the-loop. We formulate the building energy optimization problem as a Markov decision process, show how deep reinforcement learning can be used to learn energy-saving recommendations, and effectively engage occupants in energy-saving actions. recEnergy is a recommender system that learns actions with high-energy-saving potential, actively distributes recommendations to occupants in a commercial building, and utilizes feedback from the occupants to learn better energy-saving recommendations. Over a four-week user study, four different types of energy-saving recommendations were trained and learned. recEnergy improves building energy reduction from a baseline saving (passive-only strategy) of 19%-26%.

With the prevalence of smartphones, pedestrians and joggers today often walk or run while listening to music. Since they are deprived of their auditory senses that would have provided important cues to dangers, they are at a much greater risk of being hit by cars or other vehicles. In this paper, we build a wearable system that uses multichannel audio sensors embedded in a headset to help detect and locate cars from their honks, engine, and tire noises, and warn pedestrians of imminent dangers of approaching cars. We demonstrate that using a segmented architecture consisting of headset-mounted audio sensors, a front-end hardware platform that performs signal processing and feature extraction, and machine learning-based classification on a smartphone, we are able to provide early danger detection in real time, from up to 60 m away, and alert the user with low latency and high accuracy. To further reduce power consumption of the battery-powered wearable headset, we implement a custom-designed integrated circuit that is able to compute delays between multiple channels of audio with nW power consumption. A regression-based method for sound source localization, angle via polygonal regression, is proposed and used in combination with the IC to improve the granularity and robustness of localization.

Commercial buildings consume a large portion of the total electricity in the United States. One method for energy saving in commercial buildings targets inefficiencies of unoccupied spaces by relaxing the setpoint temperature. However, energy savings are severely limited when occupants are assumed to be "immovable objects"; instead, by encouraging occupant participation in the optimization, a much greater amount of energy savings can be achieved. In this work, we build on this idea and introduce energy saving recommendations based on occupant location. We introduce two types of energy saving recommendations based on location: move recommendations, which recommends the occupant to move from one space to another, and shift schedule recommendations, which recommends the occupant to arrive or depart a set amount of time earlier or later. To investigate the effects of the energy saving recommendations, we introduced a tightly coupled system composing of a simulator and a recommender system. Simulations in our building testbed revealed that energy saving recommendations coupled with occupancy-based HVAC energy management saves 25% more energy than occupancy-based HVAC energy management alone.

We propose a system that tracks each occupant’s personal share of energy use, or “energy footprint,” inside commercial building environments and provides insights to occupants on the real-time energy impact of their actions. We propose a new space-centric policy for fair apportionment of energy in shared environments and demonstrate a method for automatically determining space-centric energy zones. In this work, we design and implement ePrints, a system for tracking personalized energy usage in real-time. ePrints supports different apportionment policies, with microsecond-level footprint computation time and graceful scaling with size of building, frequency of energy updates, and rate of occupant location changes. Finally, we present applications enabled by our system, such as mobile and wearable applications to provide users timely feedback on the energy impacts of their actions, as well as applications to provide energy saving suggestions and inform building-level policies.

With the prevalence of smartphones, pedestrians and joggers today often walk or run while listening to music. Since they are deprived of their auditory senses that would have provided important cues to dangers, they are at a much greater risk of being hit by cars or other vehicles. In this demonstration we present SEUS, a wearable system aimed at Sense Enhancement for Urban Safety. SEUS uses a three-stage architecture, consisting of headset mounted audio sensors, an embedded front-end for signal processing and feature extraction, and machine learning based classification on a smartphone, to provide early danger detection for pedestrians in real-time.