Underwater activities are becoming more and more important, and underwater optical-wireless-communication (OWC) will be needed. In this work, we present and demonstrate an underwater optical-camera-communication (UWOCC) employing long-short-term-memory-neural-network (LSTM-NN). The optical transmitter (Tx) is a PMMA side-glow optical fiber, acting as an "omnidirectional optical antenna" and providing wide field-of-view (FOV) light emission. By utilizing LSTM-NN, the bit-error-rate (BER) of the UWOCC system can be improved and the data-rate is enhanced to 2.7-kbit/s, satisfying the pre-forward-error-correction (FEC) threshold.
In computer vision for aerial images, finding correct point correspondence among images plays an important role in many applications, such as image stitching, image retrieval, visual localization, etc. Most of the research works focus on the matching of local feature before a sampling method is employed, such as RANSAC, to verify initial matching results via repeated fitting of certain global transformation among the images. However, incorrect matches may still exist. Thus, a novel sampling scheme, Pentagon-Match (PMatch), is proposed in this work to verify the correctness of initially matched keypoints using pentagons randomly sampled from them. By ensuring shape and location of these pentagons are view-invariant with various evaluations of cross-ratio (CR), incorrect matches of keypoint can be identified easily with homography estimated from correctly matched pentagons. Experimental results show that highly accurate estimation of homography can be obtained efficiently for planar scenes of the HPatches dataset, based on keypoint matching results provided by LoFTR.
Foundation models represented by ChatGPT, have initiated an outstanding revolution across various domains. With the pre-trained foundation model in specific fields, numerous downstream tasks exhibit state-of-the-art performances. This paper extends this paradigm to automatic modulation classification (AMC), employing a masked autoencoder vision transformer (ViT-MAE) as a foundation model to advance AMC. The experimental results show that our signal constellation diagram-based foundation model outperforms traditional deep learning methodologies, underscoring the vast potential of foundation models in AMC and wireless communication systems.
To save power during surveillance, birds or bats can perch and rest on branches. To date, few flapping-wing drones can perch to a branch despite being capable of flying agilely in multiple flight modes. Due to their small size and limited payload, the flight endurance and surveillance times of flapping drones were short. For example, our 63-gram prototype of the hoverable flapping drone can only hover in the air for less than 5 minutes using an 11-gram lithium polymer battery. Extra loading with a camera and grippers will further reduce the endurance by 40%. The passive perching requires no flight power. This motivated our development of a 6-gram swing hook for branch perching just like the hangar hook with a lifting strength of up to 2 kilograms. With a hook onboard, the drone can perch upside down from a branch, significantly extending the surveillance duration. In addition, we outlined the principle of how pilot vision helps to aim the hook engagement to a branch by proper leveling and approaching speed. In the future, a firstperson view could further help the flight maneuver to perch indoors.
Integrated space-air-ground network (SAGIN) has become a key approach to achieve seamless global coverage and efficient information transmission. However, the requirements for task latency and the limited energy supply of unmanned aerial vehicle UAV) pose significant challenges when it acts as edge nodes to provide computation offloading services. To address the computation offloading problem in SAGIN, this paper proposes an optimization solution based on the Deep Deterministic Policy Gradient (DDPG) algorithm. We introduce behavior noise and employ state normalization to preprocess the observed state. Our aim is to minimize the energy consumption and task processing latency of UAV by jointly optimizing the UAV's flight trajectory, transmission power, task offloading ratio, and offloading destinations selection. Experimental results validate the effectiveness and feasibility of the proposed method, showing that the DDPG algorithm has significant advantages over some other intelligent algorithms in reducing energy consumption and processing latency.
