MobiCom 2025 Papers

UAV-assisted inspection is critical for modern power grid maintenance, enhancing efficiency and safety in remote areas. However, automatically designing UAV inspection trajectories is challenging due to the cluttered inspection environments, small inspection targets, and pervasive obstacles. We propose Auto-UIT, a novel method for generating inspection trajectories in noisy, sparse, and complex 3D point cloud. Auto-UIT has three core techniques: (1) A local structure-enhanced pylon segmentation, which accurately segments pylons, power lines, and surroundings in noisy point cloud for effective inspection target identification and trajectory planning. (2) A 3D fingerprint-based pylon type recognition that compensates for point cloud sparsity to complete missing inspection targets based on the pylon type. (3) An adaptive trajectory generation that samples positions in response to diverse pylon orientations and pervasive environmental obstacles, ensuring UAV operational safety. Our experiments on a real-world dataset across four distinct areas demonstrate that Auto-UIT outperforms existing baseline methods in all three tasks. Furthermore, a four-month deployment in a power grid inspection system—covering a 270 km2 primary mountainous area—yielded an expert first-review acceptance rate of 91.86% for the generated trajectories, and reduced design time by an average of 88.19% compared to manual methods, significantly improving inspection efficiency.

This paper addresses the interference challenges, aiming to improve spatial reuse and optimize spectrum efficiency in LPWANs. We reveal that existing strategies such as interference cancellation and MIMO are ill-suited to the low-cost low-rate characteristics of LPWANs. Our work introduces a novel framework, HydraNet, which leverages the capture effect of LPWAN radios to enable robust concurrent transmissions. HydraNet exempts from strict clock synchronization or accurate channel estimation as required by conventional spatial reuse strategies for interference nulling. We conduct in-depth studies with LoRa radios to uncover their underlying packet reception mechanisms and for the first time characterize their unique capture effect. Based on the new findings, we devise novel strategies to jointly control the timing and power of concurrent LPWAN transmissions. These strategies ensure sufficient power differences between packets and interference at their intended receivers. We prototype HydraNet and integrate with operational LoRaWANs and comprehensively evaluate its performance. Results show that HydraNet achieves higher spectrum utilization with up to 3.6 × throughput improvements over the state-of-the-art.

We present HyperCam, an energy-efficient image classification pipeline that enables computer vision tasks onboard low-power IoT camera systems. HyperCam leverages hyper-dimensional computing to perform training and inference efficiently on low-power microcontrollers. We implement a low-power wireless camera platform using off-the-shelf hardware and demonstrate that HyperCam can achieve an accuracy of 93.60%, 84.06%, 92.98%, and 72.79% for MNIST, Fashion-MNIST, Face Detection, and Face Identification tasks, respectively, while significantly outperforming other classifiers in resource efficiency. Specifically, it delivers inference latency of 0.08–0.27s while using 42.91–63.00KB flash memory and 22.25KB RAM at peak. Among other machine learning classifiers such as SVM, xgBoost, MicroNets, MobileNetV3, and MCUNetV3, HyperCam is the only classifier that achieves competitive accuracy while maintaining competitive memory footprint and inference latency that meets the resource requirements of low-power camera systems.

With global population growth, sustainable agriculture requires efficient soil moisture sensing for precise irrigation. While commercial soil moisture sensors are often limited by cost and durability, state-of-the-art RF-based sensing solutions require additional signal transmitter infrastructure, hindering widespread adoption. To fill this gap, we introduce GPSoil—a novel soil moisture sensing system that leverages pervasive Global Navigation Satellite System (GNSS) signals. On the hardware side, we employ two antennas along with a low-cost RF switch, enabling error mitigation for long propagation distances by comparing the signals captured from both antennas. On the software side, we leverage the inherent clock drift errors in commercial GNSS sensors and turn them into tools for fine-grained value extraction, enabling high-resolution sensing from otherwise coarse data. By integrating hardware and software innovations, we successfully achieve practical soil moisture sensing using GNSS signals. Built with off-the-shelf components, GPSoil achieves a soil moisture accuracy of 5.2% at a material cost of only $10.56, making it far more affordable than existing systems. GPSoil can sense moisture up to one meter underground, far surpassing other RF-based sensing systems and meeting the needs of most crops and irrigation systems. This work pioneers the use of GNSS signals in agriculture, offering a scalable, low-cost solution for advancing precision irrigation and promoting sustainable agriculture.

LoRa-based contact-free wireless sensing has attracted a lot of attention owing to its long sensing range, enabling wide-area sensing for the first time. While promising, existing LoRa sensing assumes there is no communication going on which is usually not true. We observe a severe degradation of sensing performance in real-world settings in the presence of communication. This issue hinders LoRa sensing from being adopted in real life and being integrated into the already established LoRa networking infrastructure. In this paper, we propose LSencom which takes the first step toward making LoRa-based wireless sensing work in the presence of communication. The key design is to employ the reversed chirp, i.e., downchirp, for sensing while keeping the original upchirp for communication. This design smartly leverages the orthogonality between downchirp and upchirp to mitigate the interference between communication and sensing. While the upchirp-downchirp design can remove most of the interference, we further adopt a novel chirp rotation method to deal with the remaining power leakage interference from upchirp to downchirp, enhancing the sensing performance. We implement LSencom on commodity LoRa nodes. Real-world experiments demonstrate that LSencom can reduce the communication-induced interference on sensing by 22.3 dB, and enable LoRa sensing even in the presence of multiple communication links.

Wi-Fi human sensing, boosted by latest progress in both system innovation and deep analytics, has demonstrated ever-increasing resolution of users' activities. Nonetheless, it may become a spy on users' private activities such as password entry or intimate social interactions. Existing countermeasures include signal obfuscation and adversarial perturbations to hamper and confuse Wi-Fi sensing, yet they both require substantial changes in Wi-Fi hardware/firmware, and they at most stay at user level in protection granularity. This paper presents Poison2Cure, the first semantic-level privacy-preserving framework for Wi-Fi human sensing systems, with full compatibility to any underlying hardware. The innovation behind Poison2Cure lies in feeding poisoned training data from (privacy-sensitive) users to the neural model for Wi-Fi sensing, degrading only the sensing for private activities while retaining that for regular ones. Moreover, we tackle the harsh conditions where the neural model is kept confidential and/or preceded by data cleansing. Our extensive evaluations demonstrate that Poison2Cure reduces over 76% of the accuracy for the private activities while keeping the accuracy for regular activities largely intact.

In-home resistance training (RT) is a convenient and effective way to maintain health and well-being. However, incorrect exercise execution can result in unintended muscle engagement and an increased risk of injury. Without access to professional coaching, an accurate muscle-aware motion feedback system becomes essential for safe and effective training. However, existing visual language models (VLMs) struggle to provide accurate and effective muscle-aware movement guidance due to their limited understanding of RT motion and the absence of related expert knowledge. In this work, we introduce Myo-Trainer, the first vision-based muscle-aware motion feedback system that uses explicit muscle-aware motion analysis and domain-specific expert knowledge to provide corrective guidance on muscle engagement and movement execution. Also, we propose a novel DAGCN-Former network that integrates both spatial and temporal modeling capabilities to capture the complex dynamics of human RT motion. Experiments involving 26 subjects and 1000+ minutes of RT demonstrate that Myo-Trainer improves the accuracy of motion analysis by 17.22%, achieves a 2.5x reduced inference latency and a BertScore of 85.88% of generated feedback compared to those provided by experienced certified trainers, outperforming existing solutions. Additionally, Myo-Trainer received higher satisfaction ratings from participants compared to other AI trainers and video tutorials, highlighting its potential for real-world applications.

Although on-device video super-resolution enables high-quality live 360-degree streaming on mobile devices, existing methods often waste energy by overlooking perceived visual quality. In this paper, we present EOS, an energy-efficient on-device super-resolution system for mobile omnidirectional video (ODV) live streaming. EOS reduces energy waste by dynamically adjusting super-resolution complexity based on the predicted visual quality of super-resolved frames. This approach raises two challenges: (1) designing an adaptive inference policy that maximizes energy savings while minimizing degradation in Quality-of-Experience (QoE), and (2) developing a method to predict visual quality under the constraints of mobile ODV live streaming. To tackle these challenges, EOS introduces EOS SR and a No-Reference Up-scaling Quality Prediction scheme. EOS SR employs a device-agnostic, scalable deep neural network optimized for mobile devices, with an energy-aware scheduler that jointly selects the optimal super-resolution model and GPU frequency. The No-Reference Upscaling Quality Prediction scheme estimates visual quality across arbitrary viewpoints in real time without requiring high-resolution reference videos. Experiments on commodity smartphones show that EOS reduces average power consumption by 34.6%–49.9% compared to baseline methods, while preserving high visual quality and frame rates.

Human behavior sensing has been receiving growing attention from both academia and industry in recent years. We report - to the best of our knowledge - the first AI-driven nationwide human behavior sensing system called SMILE in urban last-mile delivery. SMILE detects real-time human behaviors with self-supervised sensor data pretraining and uploads detection results to cloud servers with mobile networks. During its full deployment phase at JD Logistics, SMILE is deployed on over 500,000 mobile devices carried by over 300,000 delivery couriers who travel more than 10 million KM every day in 366 Chinese cities. SMILE serves the delivery of 7 billion E-commerce orders every year for more than 500 million customers. SMILE detects walking, upstairs, downstairs, still, and driving behaviors. SMILE has been fully deployed at JD Logistics to support two real-world applications that benefit both the delivery couriers and the logistics platform: (1) workload measurement to improve couriers' welfare; (2) large-scale delivery map data generation to improve the logistics delivery efficiency.

Parallelizing passive Radio Frequency Identification (RFID) reading is an arguably crucial, yet unsolved challenge in modern IoT applications. Existing approaches remain limited to time-division operations and fail to read multiple tags simultaneously. In this paper, we introduce QuinID, the first frequency-division multiple access (FDMA) RFID system to achieve fully parallel reading. We innovatively exploit the frequency selectivity of the tag antenna rather than a conventional digital FDMA, bypassing the power and circuitry constraint of RFID tags. Specifically, we delicately design the frequency-selective antenna based on surface acoustic wave (SAW) components to achieve extreme narrow-band response, so that QuinID tags (i.e., QuinTags) operate exclusively within their designated frequency bands. By carefully designing the matching network and canceling various interference, a customized QuinReader communicates simultaneously with multiple QuinTags across distinct bands. QuinID maintains high compatibility with commercial RFID systems and presents a tag cost of less than 10 cents. We implement a 5-band QuinID system and evaluate its performance under various settings. The results demonstrate a fivefold increase in read rate, reaching up to 5000 reads per second.

With the development of the Android system and increasing screen size, the use of multi-window mode has become prevalent among users. However, the security and privacy implications associated with this mode have not been thoroughly investigated. This paper uncovers severe and unique security vulnerabilities in Android's multi-window mode, revealing several high-risk side-channels that facilitate diverse cross-window attacks, leading to significant breaches of user privacy. In detail, our research introduces WinSpy, a framework leveraging a newly discovered resource contention side-channel in multi-window mode to fingerprint app launches, web pages, and in-app activities, all without violating Android's permission framework. Our extensive evaluations demonstrate that WinSpy achieves high accuracy (from 70 to 80% detecting website and app launches to over 97% recognizing critical in-app activities). Additionally, we reveal that due to Android's lenient permission management for this mode, window apps can also use Inertial Measurement Unit sensors to launch attacks, such as inferring the user's touch positions outside the window with high precision. Furthermore, we propose systematic mitigations against these vulnerabilities.

Voice assistants (VAs) have become integral to daily life, yet their always-on microphones make them attractive targets for attacks that threaten user privacy and safety. We present MetaGuardian, the first system to leverage acoustic metamaterials to defend against three major classes of attacks for VAs - inaudible, adversarial, and laser-based - within a single, portable design. Unlike prior defenses, MetaGuardian can be seamlessly integrated into the enclosures of commercial smart devices, providing strong protection without requiring software modification, hardware redesign, or costly machine learning models. MetaGuardian leverages mutual impedance effects between metamaterial units to extend the protection range to 16–40 kHz, effectively blocking wideband inaudible attacks. It also employs a carefully designed coiled space structure to disrupt adversarial signals while preserving normal VA operations. Its universal design allows flexible adaptation to different devices, striking a balance between portability and protection effectiveness. In controlled evaluations, MetaGuardian achieves a high defense success rate across all attack types, offering a practical and reliable foundation for securing VAs on smart devices.

Free-Viewpoint Videos (FVVs) enable immersive viewing of a scene from any position and angle using virtual reality (VR) head-mounted displays (HMDs), thus have great potential in various applications such as telepresence, gaming, and education. Recently, 3D Gaussian Splatting (3DGS) has emerged as a promising method for FVV construction due to its superior reconstruction quality. However, its real-time rendering on untethered HMDs remains challenging due to high computational demands. To address this challenge, we propose EdgeGaussian, a novel edge-client collaborative framework for real-time FVV rendering. Our approach employs decomposed static-dynamic 4D Gaussian splatting (SD-4DGS) to separately reconstruct static and dynamic components of a scene. We further introduce a hybrid mesh-4DGS neural representation, where static components are modeled as textured meshes for local rendering, while dynamic components are offloaded to edge servers as 4DGS. This decomposition significantly reduces the computational burden on the client device while maintaining high rendering quality. Our testbed experiments demonstrate that Edge-Gaussian achieves up to 128 FPS, outperforming state-of-the-art local rendering methods by 4x and edge rendering methods by 5x.

Wi-Fi is deemed as a promising sensing media due to its ubiquity, yet Wi-Fi sensing is known to be confined by its limited bandwidth that leads to insufficient range resolution. Though sampling a wider spectrum multiple times can enable wideband sensing, its practicality is still hampered by the need for accessing Wi-Fi firmware. In this paper, we propose μCeiver-Fi to exploit spectrum resources for fine-granularity Wi-Fi sensing; it relies solely on a commodity multi-link receiver. Since the channel samples from multiple links under the same receiver can still be misaligned, we first innovate in a comprehensive calibration process to align these samples. This is followed by a novel optimization framework to extend effective sensing bandwidth to GHz-level using only a few channel samples. Finally, we specifically design a spectral representation for sensing information in order to bridge between wideband signals and diversified downstream applications. Through comprehensive evaluations in Wi-Fi pose estimation task, we demonstrate the promising performance of μCeiver-Fi in fine-granularity sensing.

Ambient IoT (A-IoT) aims to connect hundreds of billions of ultra-low-power and battery-free devices, which has been included in the agenda for 6G standardization by 3GPP. Backscatter communication is considered the mainstream enabling technique for A-IoT; however, current state-of-the-art can hardly meet A-IoT's main technical requirements simultaneously: power consumption below 100 μW, communication ranges up to 100 m, and 100+ concurrency. This paper presents NanoScatter, the first backscatter network with each tag implemented using our customized backscatter communication ASIC. We propose a nanowatt wake-up receiver design and a sensitivity-driven downlink/uplink modulation mechanism to carry out the ASIC, which enables minimizing the tag's power consumption and long-range communication. NanoScatter supports concurrent communication of 6 IC-based tags with a subcarrier capacity of 512, achieving communication distances of 66 m indoors and 100 m outdoors. The tag consumes 1 μW in idle listening, with the core circuit using 58 nW and 43 μW during communication.

Beyond communication, wireless signals have been extensively utilized for localization, tracking, and sensing in recent years. The key information extracted for these purposes includes distance and angle. While distance measurement accuracy is mainly limited by signal bandwidth, angle accuracy depends on the number of antennas and phase noise. Conventional approaches typically improve angle estimation by boosting signal strength and increasing the number of antennas. In this paper, we propose employing a quantum receiver to substantially improve angle estimation performance. Rather than amplifying signal strength, the quantum receiver reduces the inherent hardware noise. Furthermore, we exploit the unique properties of a quantum RF receiver to construct a multi-antenna quantum system. Using only two physical quantum antennas, we generate virtual antennas by leveraging the receiver's broad frequency range, effectively increasing the number of antennas and significantly improving angle measurement performance. Our experimental results demonstrate that, with only two quantum antennas, we achieve angle estimation performance surpassing that of a conventional RF receiver equipped with 40 antennas. Furthermore, quantum antennas are not constrained by the coupling effects that typically limit the spacing between conventional RF antennas, allowing for much closer placement. This represents a significant step toward reducing the size of antenna arrays while preserving localization and tracking performance.

This paper presents Serafin, the first full-stack, sub-mW, and versatile Bluetooth-enabled RF sensor that brings transparent RF sensing to any mobile and IoT device: it independently conducts the whole RF sensing process from RF signal reception to sensing result computation in a wide variety of sensing tasks with only negligible power consumption. At the core of Serafin are our two designs that address the challenge posed by the stringent sub-mW power constraint to jointly achieving versatility and full stackness. Specifically, (i) we utilize the ambient Bluetooth advertising signal as the signal for sensing, and extract the phase difference of the sensing signals received by each antenna pair from the amplitude of their sum signal, which avoids power-hungry hardware components and intensive computation, and (ii) we employ low-power MCU as the computation hardware, and suppress its power consumption by activating it adaptively only when necessary and customizing a light-weight neural network model that still ensures satisfactory inference accuracy. Our extensive experiments on 6 representative sensing tasks show that Serafin achieves competitive sensing performance, but consumes only around 500–900μW power, which is 3–4 orders of magnitude lower than those of existing full-stack and versatile counterparts.

LoRa, as a widely used Low-Power Wide-Area Network (LP-WAN) technology, is designed for long-term use. However, in practice, the battery life of LoRa devices often falls far short of the expected decade-long duration. We propose OptSamp, a software-based approach that reduces the energy consumption of LoRa devices by lowering their physical-layer sampling rate, thereby extending battery life. To address the challenge of frequency aliasing caused by down-sampling, we embed a specially designed feature into each modulated symbol, enabling the OptSamp receiver to accurately recover the distorted signal. In addition, we design an adaptive sampling-rate selection mechanism to balance link reliability and energy efficiency. We further propose OptSamp+, which compresses symbol duration to shorten uplink transmission time, thereby reducing transmitter energy consumption and enhancing spectral efficiency. We prototype OptSamp and OptSamp+ on software-defined LoRa platforms and evaluate them in both indoor and outdoor environments. Results show that OptSamp reduces the receiver-side sampling rate to 1/16 of the Nyquist rate, cutting downlink energy consumption by 68%, and OptSamp+ reduces uplink transmission time by up to 1/32 compared to traditional LoRa.

Diffractive neural networks (NNs) have garnered attention for directly implementing wireless signal processing at the physical layer. However, they are limited by a constrained weight learning space and activation functions, which restricts their data processing capabilities. To address this, we propose an RF circuit-based weighted diffraction NN (WDNN) that rivals digital NNs in processing ability. We design a weighted asymmetric RF coupler unit that, when stacked into a network, enables diffractive propagation with arbitrary connection weights. Additionally, an activation module is introduced that utilizes RF amplifiers operating in their nonlinear regions. We validate the effectiveness of the proposed WDNN through three tasks: 32-level amplitude modulated (AM) signal decoding, 31-class angle of arrival (AoA) estimation, and 2-class Wi-Fi based fall detection. After training, WDNN achieves the accuracy of 98.5%, 93.7%, and 90.8% in the AM decoding, AoA estimation, and fall detection tasks, respectively; while the diffractive NN SOTA achieves only 21.6%, 16.9%, and 63.3%. We also implement the prototypes of WDNN and SOTA, and real-world experimental results demonstrate that our method achieves an average accuracy improvement of up to 76.85% across various tasks compared to SOTA.