![Robotic Computing on FPGAs](/uploads/posts/2021-07/1625757409_cov250mk.jpg)
Автор: Shaoshan Liu, Zishen Wan, Bo Yu
Издательство: Morgan & Claypool
Год: 2021
Страниц: 220
Язык: английский
Формат: pdf (true)
Размер: 16.7 MB
This book provides a thorough overview of the state-of-the-art field-programmable gate array (FPGA)-based robotic computing accelerator designs and summarizes their adopted optimized techniques. This book consists of ten chapters, delving into the details of how FPGAs have been utilized in robotic perception, localization, planning, and multi-robot collaboration tasks. In addition to individual robotic tasks, this book provides detailed descriptions of how FPGAs have been used in robotic products, including commercial autonomous vehicles and space exploration robots.
The authors believe that FPGAs are the best compute substrate for robotic applications for several reasons. First, robotic algorithms are still evolving rapidly, and thus any ASIC-based accelerators will be months or even years behind the state-of-the-art algorithms. On the other hand, FPGAs can be dynamically updated as needed. Second, robotic workloads are highly diverse, thus it is difficult for any ASIC-based robotic computing accelerator to reach economies of scale in the near future. On the other hand, FPGAs are a cost effective and energy-effective alternative before one type of accelerator reaches economies of scale. Third, compared to systems on a chip (SoCs) that have reached economies of scale, e.g., mobile SoCs, FPGAs deliver a significant performance advantage. Fourth, partial reconfiguration allows multiple robotic workloads to time-share an FPGA, thus allowing one chip to serve multiple applications, leading to overall cost and energy reduction.
Specifically, FPGAs require little power and are often built into small systems with less memory. They have the ability of massively parallel computations and to make use of the properties of perception (e.g., stereo matching), localization (e.g., simultaneous localization and mapping (SLAM)), and planning (e.g., graph search) kernels to remove additional logic so as to simplify the end-to-end system implementation. Taking into account hardware characteristics, several algorithms are proposed which can be run in a hardware-friendly way and achieve similar software performance. Therefore, FPGAs are possible to meet real-time requirements while achieving high energy efficiency compared to central processing units (CPUs) and graphics processing units (GPUs). In addition, unlike the application-specific integrated circuit (ASIC) counterparts, FPGA technologies provide the flexibility of on-site programming and re-programming without going through re-fabrication with a modified design.
Cameras are widely used in intelligent robot systems because of their lightweight and rich information for perception. Cameras can be used to complete a variety of basic tasks of intelligent robots, such as visual odometry (VO), place recognition, object detection, and recognition. With the development of convolutional neural networks (CNNs), we can reconstruct the depth and pose with the absolute scale directly from a monocular camera, making monocular VO more robust and efficient. CNNs have become the core component in various kinds of robots. However, since neural networks (NNs) are computationally intensive, Deep Learning (DL) models are often the performance bottleneck in robots.
Скачать Robotic Computing on FPGAs
![](/templates/Mirlib/images/MirLib.png)
[related-news] [/related-news]
Комментарии 0
Комментариев пока нет. Стань первым!