Wildfires represent an escalating global threat to both the environment and society due to their frequency, duration, and expansion, which is exacerbated by climate change. Furthermore, recent heat waves underscore the critical need for swift detection and response mechanisms to curb wildfires from escalating into fully developed and uncontrollable stages. In that sense, this manuscript addresses an aerial smoke detector, composed of an unmanned aerial vehicle equipped with a camera, where visual information is sent to a ground station based on robotic operating system, where the proposed smoke detection methodology is deployed in real time. Such approach is formulated on an optimized YOLOv8 nano model, which is specifically trained using a customized data base with smoke under defying conditions. This solution ensures peak performance even within limited computational resources. The experimentally tests conducted first using images and videos, then, taking the video from the drone under controlled laboratory conditions, and finally by unstructured field experiments, such scenarios determine its robustness under such challenged conditions, producing confidence over 70%, and reducing the bias by validation metrics such as 95% of precision, and 88.5% of recall, respectively.

The increasing use of composite materials in sectors like automotive and aerospace poses serious problems for preserving their material performance and integrity. Because they provide automated, accurate, and effective inspection capabilities, advanced inspection techniques—in particular, robotic intelligence technologies—have emerged as viable options. This paper provides a comprehensive review of the key robotic intelligence technologies used in the inspection of composite materials, highlighting advancements in vision-based, tactile-based, and force-based traditional approaches, as well as the development in modern advanced deep learning methods such as Convolutional Neural Network (CNN) based image processing techniques for inspection. In order to guarantee accurate and steady manipulation during inspection jobs, robot control strategies are also investigated. The robot’s capacity to navigate intricate composite constructions while preserving constant inspection quality has also been greatly improved by the use of clever path-planning algorithms. The paper concludes by outlining future directions for improving inspection accuracy and efficiency through AI integration and advanced sensor technologies.

Scene flow estimation provides the 3D low-level motion understanding in dynamic scenes. In this paper, we propose an optimization-based scene flow estimation method with neural rigidity prior for the autonomous driving environment. Specifically, we utilize the rigidity prior of dynamic scenes to partition the point clouds into pillars of different resolutions. Then, the flow vector of a point is represented as the average of local rigid transformations associated with the different pillars to which it belongs. To model local rigidity, we employ the neural implicit representation for encoding the intrinsic constraints of pillars. Our method achieves state-of-the-art accuracy on three commonly-used autonomous driving datasets: Argoverse, Waymo, and nuScenes, and even surpasses previous supervised learning-based methods. Experiment results demonstrate the effectiveness of our method, particularly on sparse points in the autonomous driving scene.

Utilising thermal imaging for simultaneous localisation and mapping has effectively improved the performance and robustness of robots and autonomous systems in unconventional environments. However, the transmission of large amounts of visual data from terminal devices to the central system for training not only results in high communication costs and pressure on bandwidth, but also induces concerns regarding privacy. Meanwhile, for applications in the real world, it is essential to expand the input domain to more practical scenarios rather than relying on experimental environments, and the terminal devices in-service can also benefit from further training with data collected in operations. To deal with these challenges, we investigated FTI-SLAM by applying federated learning to a thermal-inertial simultaneous localisation and mapping system. We conducted a series of experiments and showed that federated learning is feasible for the task and can improve overall performance.

Path planning algorithms are indispensable for controlling micro-/nanorobots through complex and unknown environments in the biomedical and medical fields. With the tasks performed becoming more complex, higher-quality paths are required to avoid obstacles for ensuring the safe and efficient movement of micro-/nanorobots. A comparative analysis of path planning algorithms is conducted to elucidate the algorithm’s application and optimization for different environments. According to the environment modeling approach, existing path planning algorithms are classified into searching, sampling, and dynamic aspects. Searching path planning algorithms directly retrieve the global path possessing minimum cost from the modeled static waypoints. Sampling path planning algorithms employ randomly sampled waypoints within the target space, which eliminates the necessity for environmental modeling. Dynamic path planning algorithms utilize local paths to regulate the motion of micro-/nanorobots in real time. Deep learning networks based on big data will become an important research direction for the control and navigation of micro-/nanorobots. The advantages and limitations of path planning algorithms in varied spatial contexts are elucidated through detailed examples and descriptions, providing a comprehensive understanding of performance and applicability. This review underscores recent advancements in this emerging domain and stands as a testament to the dynamic landscape of micro-/nanorobotics and the continual pursuit of superior motion control solutions.