We study a coordinated delivery problem where a drone, operating in tandem with a truck traveling along a road, delivers tools to pylons on a high-voltage power line. Each delivery is defined by a target pylon, package weight, urgency-based profit, and feasible launch and rendezvous points. The drone faces payload and energy constraints, can serve multiple deliveries in a mission, and can perform several missions before recharging. The objective is to maximize total urgency profit across all executed missions. We formalize this as the Power line Maintenance Problem (PMP), which requires selecting a subset of feasible missions that jointly respect energy limits and individually satisfy payload capacity. We prove that PMP is NP -hard and propose a two-phase solution: Generate Missions Phase (GMP), which generates multi-delivery missions, and Truck-drone multi-package Mission Problem (TMP), which selects the most profitable subset. An Integer Linear Program (ILP) formulation is provided for computing optimal solutions. We further study four variants of PMP with different weight models and energy costs. Complexity analysis shows that GMP is polynomial or pseudo-polynomial, while TMP remains NP -hard. Experiments show that GMP greatly reduces candidate missions, while allowing multiple deliveries per mission improves energy efficiency by up to 4 ×.
In an era dominated by data-driven solutions, Deep Neural Networks (DNNs), which have been proven to be pivotal tools in extensive applications across various domains, are evolving in terms of both depth and architecture to meet the escalating demands of contemporary utilizations. Nevertheless, deploying a complex DNN model on mobile devices may result in substantial processing latency and increasing energy consumption. The emerging Mobile Edge Computing (MEC), characterized by the allocation of computing capacity at the access point, enables the partitioning of DNN models so as to conserve energy on mobile devices and mitigate inference latency. Existing DNN partitioning methods typically train prediction models offline to make partition decisions and reduce end-to-end inference latency, which requires a great number of labeled datasets and may incur a prolonged pre-processing duration. In this paper, we develop an online Deep Reinforcement Learning (DRL) based adaptive partition method to dynamically determine optimal partitioning decision so as to jointly accelerate DNN inference and mitigate energy consumption. We run the proposed algorithm in an edge computing scenario consisting of NVIDIA Jetson Nano and an edge server equipped with RTX3090 for four different DNN models, including VGG16, MobileNetV2, ResNet50 and GPT2-medium. Then we collect actual processing latency and energy consumption and compare the performance of the proposed algorithm with state-of-the-art solutions. The experimental results demonstrate that, even under varying channel conditions, DAPart can achieve an average reduction of 38.8% in latency and 36.5% in energy consumption compared with other available methods.
As one widely used wireless technique, WiFi has the potential to perform non-contact monitoring of respiration. While pioneering works have demonstrated promising performance, they inevitably ignore the impact of pervasive motion interference caused by non-target body areas or nearby surrounding interferers. Furthermore, motion interference has a significant negative impact on the accuracy of respiration monitoring due to signal aliasing, which further decays the recognition accuracy of these systems. To tackle these limitations, by employing a novel beamforming algorithm, we propose Wi-Spatial , an anti-motion interference respiration monitoring system based on commodity WiFi devices that can robustly mitigate the motion interference of non-target persons and further effectively extract the respiration signals from the target person. Wi-Spatial first enables stable beam pattern generation of channel state information (CSI) via adaptive phase shift compensation. Then, a directional nulling beamforming (DN-BF) scheme that directs the main lobe toward the target subject while adaptively placing beam nulls in the directions corresponding to the interferer is proposed to enhance the signal response in the area where the target subject is located. Extensive experiments have demonstrated that Wi-Spatial achieves high performance in monitoring respiration under various intensity motion inferences, especially in high-intensity motion interference, with a median absolute deviation of 0.41 bpm, which outperforms the state-of-the-art solutions by \(28\% \) .
Cooperative perception enables connected autonomous vehicles to extend their sensing range and overcome occlusions by exchanging sensor data. However, its real-world deployment is hindered by asynchronous sensor streams and inaccurate localization of occluded regions. This work presents RAO++ , a real-time cooperative perception system that merges asynchronous sensor data from different vehicles through our novel designs of motion-compensated occupancy flow prediction , on-demand data sharing , with a variety of system optimizations to improve the accuracy and coverage of the perception system. Our comprehensive evaluation, including real-world and emulation experiments under diverse LiDAR configurations, shows that RAO++ outperforms asynchronous-unaware methods by more than 34% in perception coverage and by up to 14% in perception accuracy. Moreover, RAO++ reduces latency by 1.2–3.5× and communication overhead by 45–70% compared with the state-of-the-art asynchronous-aware baseline, while maintaining comparable detection accuracy. Finally, RAO++ demonstrates a practical data overhead of 12.8 KB per frame, enabling deployment under realistic bandwidth constraints.
Drinking water temperature is linked to gastrointestinal diseases such as gastroesophageal reflux disease, esophageal cancer, and functional dyspepsia. This paper presents HearT , a headphone-deployable system that leverages fluid dynamics–based vibration modeling to infer the temperature of ingested fluids from bone-conducted swallowing sounds. The system first examines the stability and dominant source components of swallowing across different phases. It then models temperature-dependent vibration parameters for each source and maps acoustic features to these parameters. This physics-driven framework achieves strong performance even with classical machine learning models. HearT attains average errors lower than typical human subjective estimation and the oral thermal perception threshold. Compared with contact-based temperature measurements or ambient sensing, HearT enables continuous tracking of local temperature fluctuations within the body’s broader thermal field, while benefiting from the accessibility, robustness, and comfort of commercial headphones.
