其他名称:
24-71-00024
项目负责人:
Kalita Diana
发表日期:
2024
主持机构:
North-Caucasus Federal University,
国家:
俄罗斯
开始日期:
2024
结束日期:
2026
简介:
In the current era of autonomous robotics, unmanned ground vehicles and unmanned aerial vehicles have become the main technology for implementing security and surveillance systems. In addition to their first military applications, UGVs and UAVs are finding their way into remote monitoring and data collection. Specially designed UAVs find their way into photography, inspection and surveillance missions and other applications. Autonomous ground vehicles open up new opportunities for smart roads, intelligent road safety and traveler comfort, and the interest in research on autonomous driving algorithms is due to the significant ability to reduce accidents and human casualties. Researchers believe that more than 8 million drones will be used by 2025, which confirms the relevance of introducing advanced technological solutions into autonomous unmanned ground and air transport systems.$In both cases, accuracy, latency, energy efficiency, and economy are key requirements for autonomous vehicle computing systems. Accuracy and latency affect security, while power efficiency affects autonomy. Currently, a large number of studies are aimed at improving the accuracy of the algorithms necessary to achieve the level of system automation, however, with an increase in accuracy, there is a reverse increase in speed and energy efficiency indicators. Solving problems aimed at improving not only the accuracy of the operation of unmanned transport computing systems, but also increasing speed while reducing hardware costs that characterize the efficiency of the system will significantly expand the use of unmanned ground and aircraft vehicles in human practice.$The search for advanced solutions to obtain the stated requirements from computing systems requires a change in processing approaches and computational algorithms. Probabilistic filtering is a promising approach where there is uncertainty associated with the location of a moving object. In addition, probabilistic filter models are applicable as a means of pre-processing video data in the case where data denoising is required. The use of an alternative as a number system for performing arithmetic operations during such processing of video data will increase the speed of the computing system. Residua number system (RNS) is one of the well-known non-positional systems capable of performing parallelization of calculations by working with numbers of small capacity. The use of RNS will not only increase the speed of calculations, but also provide economical processing in the context of reducing energy and hardware costs.$The work under this project will be aimed at developing high-quality hardware implementations of video processing digital filter models on FPGA and ASIC computing platforms using modular computing that combines high accuracy and speed of the device with low power consumption and hardware costs to increase the availability of unmanned video processing tools vehicle.$The research planned within the framework of the project will be used to solve applied problems of video processing. For each application, a problem-oriented setting of digital filters in the RNS is assumed, depending on the conditions of the problem being solved, in order to achieve the best indicators of filtering quality, performance, power consumption and hardware costs. The current level of development of ASIC and FPGA technologies shows the ability to implement large-scale and complex projects in the field of computer vision based on it, which indicates the possibility of successful implementation of the ideas proposed in this project.
专业领域:
信息技术
语种:
英语