Zhaohui Wang

Air-ground collaborative systems (AGCS), in which unmanned aerial vehicles survey from above while ground vehicles navigate the terrain below, can achieve capabilities far beyond those of any single platform when they share perception and coordinate action. AGCS are undergoing a three-stage paradigm shift: from rule- and reinforcement learning-driven programmed execution, through vision–language navigation that introduces a language interface, to MLLM-driven autonomous perception, reasoning, and heterogeneous collaboration. Taking this paradigm shift as its organizing thread, this review critically examines how large language models (LLMs) and multimodal large language models (MLLMs) are reshaping the intelligence paradigm of air-ground heterogeneous collaborative systems between 2023 and 2026. We argue that the core of this transformation is not the substitution of one technical tool for another, but a change in how AGCS understand tasks, coordinate heterogeneous agents, and reason in unknown environments. They are moving from reliance on pre-specified structures toward semantic understanding and autonomous decision-making. Organized around a three-stage framework, the review covers collaborative perception, vision–language navigation (VLN), vision–language–action (VLA) models, multi-agent collaboration, and air-ground heterogeneous MLLM coordination. We analyze the capability boundaries of MLLMs in air-ground settings and outline the key challenges and roadmap from proof-of-concept to generalizable frameworks.

Fall detection is a vital task in health monitoring, as it allows the system to trigger an alert and therefore enabling faster interventions when a person experiences a fall. Although most previous approaches rely on standard RGB video data, such detailed appearance-aware monitoring poses significant privacy concerns. Depth sensors, on the other hand, are better at preserving privacy as they merely capture the distance of objects from the sensor or camera, omitting color and texture information. In this paper, we introduce a privacy-supporting solution that makes the RGB-trained model applicable in depth domain and utilizes depth data at test time for fall detection. To achieve cross-modal fall detection, we present an unsupervised RGB to Depth (RGB2Depth) cross-modal domain adaptation approach that leverages labelled RGB data and unlabelled depth data during training. Our proposed pipeline incorporates an intermediate domain module for feature bridging, modality adversarial loss for modality discrimination, classification loss for pseudo-labeled depth data and labeled source data, triplet loss that considers both source and target domains, and a novel adaptive loss weight adjustment method for improved coordination among various losses. Our approach achieves state-of-the-art results in the unsupervised RGB2Depth domain adaptation task for fall detection. Code is available at https://github.com/1015206533/privacy_supporting_fall_detection.

Changepoints are abrupt variations in the underlying distribution of data. Detecting changes in a data stream is an important problem with many applications. In this paper, we are interested in changepoint detection algorithm which operates in an online setting in the sense that both its storage requirements and worst-case computational complexity per observation are independent of the number of previous observations. We propose an online changepoint detection algorithm for both univariate and multivariate data which compares favorably with offline changepoint detection algorithms while also operating in a strictly more constrained computational model. In addition, we present a simple online hyperparameter auto tuning technique for these algorithms.