Non-intrusive load monitoring (NILM) implements energy decomposition in a non-intrusive manner and provides a promising path to tap demand response potential for flexible load resources in residential and commercial b...
详细信息
Unmanned aerial vehicle technology is growing rapidly as it finds its application in the various industries, including military & defense, agriculture, logistics, transportation, healthcare, entertainment and many...
Unmanned aerial vehicle technology is growing rapidly as it finds its application in the various industries, including military & defense, agriculture, logistics, transportation, healthcare, entertainment and many others. One of the fastest growing industries including drones is the entertainment industry. More specifically, First Person View (FPV) piloting has become a popular sport which attracts huge masses. In FPV systems, the pilot wears FPV goggles and controls the drone using a controller, and the drone transmits video data to the goggles in real time. Since drones usually fly very quickly, the video quality in terms of resolution, compression artefacts and end-to-end delay must be maintained. This paper explores the idea of optimizing the Motion Estimation (ME) algorithm of existing High Efficiency Video Coding (HEVC) algorithms by utilizing user input from the controller. For example, if the drone is directed to hover to the left, then it would make sense to search for the most similar blocks in the previous video frame only on the left side of the referent block. For the purpose of this research, a HEVC video coder was customized to receive additional input besides the video data – the user input from the controller, or in other words, drone movement directions. We call this ME algorithm User Input Search (UIS). Our UIS algorithm is compared with the standard ME algorithms and its efficiency is tested.
In modern industrial cyber-physical systems, a mass of process variables has been obtained by the high-sampling online sensors. Meanwhile, the key quality indexes are usually obtained infrequently from the laboratory....
详细信息
Insulator exposed to a wild environment for long-term running usually suffers from the damage of natural calamities that makes them become one of the most fragile components in the whole power system. In this case, ev...
详细信息
Globally, the circular efficiency of biomass resources has become a priority due to the depletion and negative environmental impacts of fossil fuels. This study aimed to quantify the atmosphere-dependent combustion of...
详细信息
The multiple attribute decision making (MADM) is a one of most crucial topic in decision making and computer science. The key technology for MADM is to learn the correlation between different attributes, and the graph...
详细信息
To solve the limitations of current dissolved gas analysis based transformer fault diagnosis methods and further improve the diagnostic accuracy, this paper develops a novel transformer fault diagnosis approach based ...
详细信息
As wind farms continue to grow in size, minimizing wake losses between turbines in layout design becomes increasingly important. Current wind farm layout strategies often focus on the performance of individual turbine...
详细信息
Bipolar objects are widespread in nature, and they describe the opposition and unity of the things. Inspired by decision making characterizing in terms of fuzzy graph structures, we propose a bipolar fuzzy graph-based...
详细信息
Recent advances in Machine Learning (ML) brought several advantages also within computer network management. For programmable data planes, however, it is more challenging to benefit from these advantages, given their ...
详细信息
ISBN:
(数字)9798350351255
ISBN:
(纸本)9798350351262
Recent advances in Machine Learning (ML) brought several advantages also within computer network management. For programmable data planes, however, it is more challenging to benefit from these advantages, given their limited resource capabilities colliding with the complexity of ML models. In this paper, we propose ART, an attempt to simplify ML-based solutions for routing, so that they can "fit", i.e., be executed, on P4 switches. To provide such model simplification, ART relies on efficient knowledge distillation techniques, converting, in particular, Deep Reinforcement Learning (DRL) models into a simpler Decision Tree (DT). Our evaluation results validate the accuracy of the extracted model and the application of the model logic directly into switches with little impact, paving the way for a more reactive data plane programmability via machine learning integration.
暂无评论