Anticipating opponents’ intentions in multi-agent systems is critical for rapid, robust decision-making in domains from autonomous unmanned aerial vehicle (UAV) coordination to competitive strategy games. However, most existing methods rely on unrealistic access to private opponent information or fail to capture the recursive reasoning humans use to adapt in dynamic, partially observable environments. We address these gaps with a hierarchical world-opponent modeling framework that unifies environment dynamics prediction and intention-strategy reasoning in a single architecture, without requiring private data. Inspired by human social inference, our method uses multiple learnable intention and strategy queries over local observations to recursively update opponent models, anticipate future trajectories, and adapt strategies in real time. Joint optimization of the world and opponent models captures the mutual influence between environment transitions, intentions, and maneuvers, yielding sample-efficient learning. Across benchmarks, including close-range multi-UAV engagements and the StarCraft Multi-Agent Challenge, our approach achieves up to 5.3 times faster learning than model-free multi-agent reinforcement learning baselines, while consistently improving maneuver effectiveness and decision intelligence. These results demonstrate a scalable, high-efficiency solution for adversarial reasoning in complex multi-agent cooperative-competitive settings.

The growing demand for real-time DNN applications on edge devices necessitates faster inference of increasingly complex models. Although many devices include specialized accelerators (e.g., mobile GPUs), dynamic control-flow operators and unsupported kernels often fall back to CPU execution. Existing frameworks handle these fallbacks poorly, leaving CPU cores idle and causing high latency and memory spikes. We introduce Parallax, a framework that accelerates mobile DNN inference without model refactoring or custom operator implementations. Parallax first partitions the computation DAG to expose parallelism, then employs branch-aware memory management with dedicated arenas and buffer reuse to reduce runtime footprint. An adaptive scheduler executes branches according to device memory constraints; meanwhile, fine-grained subgraph control enables heterogeneous inference of dynamic models. By evaluating on representative DNNs across different mobile devices, Parallax achieves up to 46% latency reduction, maintains controlled memory overhead (26.5%), and delivers up to 30% energy savings compared with the state-of-the-art frameworks, offering improvements aligned with the responsiveness demands of real-time mobile inference.

Recent progress on Remote Sensing Foundation Models (RSFMs) aims toward universal representations for Earth observation imagery. However, current efforts often scale up in size significantly without addressing efficiency constraints critical for real-world applications (e.g., onboard processing, rapid disaster response) or treat multispectral (MS) data as generic imagery, overlooking valuable physical priors. We introduce PhySwin, a foundation model for MS data that integrates physical priors with computational efficiency. PhySwin combines three innovations: (i) physics-informed pretraining objectives leveraging radiometric constraints to enhance feature learning; (ii) an efficient MixMAE formulation tailored to SwinV2 for low-FLOP, scalable pretraining; and (iii) token-efficient spectral embedding to retain spectral detail without increasing token counts. Pretrained on over 1M Sentinel-2 tiles, PhySwin achieves SOTA results (+1.32\% mIoU segmentation, +0.80\% F1 change detection) while reducing inference latency by up to 14.4$\times$ and computational complexity by up to 43.6$\times$ compared to ViT-based RSFMs.

In this study, we present ML-Track, an innovative uncooperative passive tracking system leveraging WiFi communication signals between multiple devices. Our approach is realized with three pivotal techniques. First, we introduce a novel protocol termed multi-link round-trip CSI, which enables multi-link bistatic Doppler detection within a WiFi network. Second, a phase error cancellation method is developed, and we demonstrate a 0.92 rad reduction in error (0.96 to 0.04 rad) experimentally. Lastly, we propose a particle-filter-based back-end to track a moving human in the room passively without the need for the participant to carry any type of cooperative or active device. A prototype system is constructed using four Raspberry Pi CM4 units and subjected to real-world evaluations. Experimental results indicate a median error of approximately 0.23 m for tracking, which corresponds to a relative error of 5.8% based on the 4 m side length of the experimental field. Compared to existing studies, a distinct advantage of our system is it can run with non-MIMO (single-antenna) WiFi devices, making it particularly suitable for budget or low-profile WiFi hardware. This compatibility makes it an ideal fit for real-world Internet-of-Things (IoT) devices. Moreover, in terms of computational demands, our solution excels, delivering real-time performance on the Raspberry Pi CM4 while utilizing just 20% of its CPU capability and drawing a modest 2.5 watts of power.

Indoor localization using WiFi signals has been studied since the emergence of WiFi communication. This paper presents a novel training-free approach to indoor localization using a customized WiFi protocol for data collection and a factor graph-based back-end for localization. The protocol measures the round-trip phase, which is very sensitive to small changes in displacement. This is because the sub-wavelength displacements introduce significant phase changes in WiFi signal. However, the phase cannot provide absolute range information due to angle wrap. Consequently, it can only be used for relative distance (displacement) measurement. By tracking the round-trip phase over time and unwrapping it, a relative distance measurement can be realized and achieve a mean absolute error (MAE) of 0.06m. For 2-D localization, factor graph optimization is applied to the round-trip phase measurements between the STA (station) and four APs (access points). Experiments show the proposed concept can offer a decimeter-level (0.26m MAE and 0.24m 50%CDF) performance for real-world indoor localization.

This paper presents a wireless Doppler sensing system using WiFi round-trip channel state information (RTCSI), which is implemented using the channel state information (CSI) from the Raspberry Pi CM4 onboard WiFi chip and a customized WiFi protocol. Utilizing the CSI phase in WiFi sensing is challenging as hardware asynchronization introduces significant phase errors. Similar to WiFi round-trip time (RTT) ranging, RTCSI was proposed to cancel the adverse effect of asynchronization through two-way communication. However, previous work mainly focuses on measuring RTCSI over frequency (different WiFi channels) to simulate a wide-band ranging. In this work, RTCSI is measured over time and a Doppler sensing prototype is built to detect a moving target in the wireless channel. Our findings show that this RTCSI-based Doppler sensing system is sensitive and effective in the real world. Moreover, it may be integrated with other techniques further to improve the performance in joint communications and sensing.

Micro-Doppler signatures contain considerable information about target dynamics. However, the radar sensing systems are easily affected by noisy surroundings, resulting in uninterpretable motion patterns on the micro-Doppler spectrogram. Meanwhile, radar returns often suffer from multipath, clutter and interference. These issues lead to difficulty in, for example motion feature extraction, activity classification using micro Doppler signatures ($\mu$-DS), etc. In this paper, we propose a latent feature-wise mapping strategy, called Feature Mapping Network (FMNet), to transform measured spectrograms so that they more closely resemble the output from a simulation under the same conditions. Based on measured spectrogram and the matched simulated data, our framework contains three parts: an Encoder which is used to extract latent representations/features, a Decoder outputs reconstructed spectrogram according to the latent features, and a Discriminator minimizes the distance of latent features of measured and simulated data. We demonstrate the FMNet with six activities data and two experimental scenarios, and final results show strong enhanced patterns and can keep actual motion information to the greatest extent. On the other hand, we also propose a novel idea which trains a classifier with only simulated data and predicts new measured samples after cleaning them up with the FMNet. From final classification results, we can see significant improvements.

The use of Wi-Fi signals for human sensing has gained significant interest over the past decade. Such techniques provide affordable and reliable solutions for healthcare-focused events such as vital sign detection, prevention of falls and long-term monitoring of chronic diseases, among others. Currently, there are two major approaches for Wi-Fi sensing: (1) passive Wi-Fi radar (PWR) which uses well established techniques from bistatic radar, and channel state information (CSI) based wireless sensing (SENS) which exploits human-induced variations in the communication channel between a pair of transmitter and receiver. However, there has not been a comprehensive study to understand and compare the differences in terms of effectiveness and limitations in real-world deployment. In this paper, we present the fundamentals of the two systems with associated methodologies and signal processing. A thorough measurement campaign was carried out to evaluate the human activity detection performance of both systems. Experimental results show that SENS system provides better detection performance in a line-of-sight (LoS) condition, whereas PWR system performs better in a non-LoS (NLoS) setting. Furthermore, based on our findings, we recommend that future Wi-Fi sensing applications should leverage the advantages from both PWR and SENS systems.

Human sensing using WiFi signal transmissions is attracting significant attention for future applications in e-healthcare, security, and the Internet of Things (IoT). The majority of WiFi sensing systems are based around processing of channel state information (CSI) data which originates from commodity WiFi access points (APs) that have been primed to transmit high data-rate signals with high repetition frequencies. However, in reality, WiFi APs do not transmit in such a continuous uninterrupted fashion, especially when there are no users on the communication network. To this end, we have developed a passive WiFi radar system for human sensing which exploits WiFi signals irrespective of whether the WiFi AP is transmitting continuous high data-rate Orthogonal Frequency-Division Multiplexing (OFDM) signals, or periodic WiFi beacon signals while in an idle status (no users on the WiFi network). In a data transmission phase, we employ the standard cross ambiguity function (CAF) processing to extract Doppler information relating to the target, while a modified version is used for lower data-rate signals. In addition, we investigate the utility of an external device that has been developed to stimulate idle WiFi APs to transmit usable signals without requiring any type of user authentication on the WiFi network. In this article, we present experimental data which verifies our proposed methods for using any type of signal transmission from a standalone WiFi device, and demonstrate the capability for human activity sensing.