A complex mechanical device capable of violent bouncing? A variety of expensive sensors? Or like the EVE to conquer the world with a cute shape?
No, no, no! The most important thing is: to complete specific work in a complex environment, to be a robot that is good for the people (good three!)
In order to achieve this lofty goal, the robot must look at the six roads, listen to all directions, walk the world, and have a chest. And all of this is inseparable from a powerful brain - high-performance computing platforms and efficient algorithms.
General-purpose processors are increasingly difficult to meet the ever-larger and more diverse data processing needs in terms of computing performance and energy efficiency. With the heterogeneous computing platform composed of coprocessors and CPUs such as GPU, FPGA and other intelligent engines to improve computing performance, it has become a research hotspot in the academic and industrial circles. As a special parallel computing method, heterogeneous computing can assign different computing tasks according to the structural characteristics of each computing subsystem, and it can not reflect the traditional architecture in terms of improving computing performance, energy efficiency ratio and real-time security. The advantages are gradually applied in various occasions where the calculation demand is large. For data center scenarios, Intel has announced the implementation of heterogeneous computing through the Xeon+FPGA platform and the Xeon Phi family of products.
So, what is the situation in the field of robots we care about?
First, what kind of computing platform does the robot need?The robot is a complex mechatronic device that combines the latest research from mechanical and precision machinery, microelectronics and computers, automatic control and drive, sensors and information processing, and artificial intelligence.
In order to integrate these results, the ability to truly form cognition, feelings and actions enables robots to understand and respond to the real world and accurately perform operational tasks. The computing platform on which they are equipped must meet a variety of requirements, including:
1) Perception ability: It has rich I/O interface and supports various I/O protocols such as USB 3, UART and I2C to receive various sensor data.
2) Control ability: It can control various moving parts, such as chassis, arm, finger and head, to complete multiple dimensions of action, and has real-time and security guarantees.
3) Computational ability: It has to deal with the processing overhead of a large amount of data, but also meet the huge computing requirements of various intelligent algorithms.
4) Power consumption control: To meet the above capabilities while minimizing power consumption, avoid becoming a battery killer.
5) Ease of use: A friendly programming model must be provided to make the application development of the robot fast and accurate.
The existing computing platform is difficult to meet the above requirements, and it is extremely urgent to build a heterogeneous computing platform for robots! The CPU+FPGA solution has come to our attention with its unique advantages.
Second , what is FPGASince its inception in 1985, FPGA-based programmable logic devices have gained a place in communications, medical, industrial, and security fields with their strengths in performance, time-to-market, cost, stability, and long-term maintenance. Especially in the past two years, with the prosperity of cloud computing, high-performance computing and artificial intelligence, FPGAs with inherent advantages have received unprecedented attention.
The FPGA consists of six parts: a programmable input/output unit, a basic programmable logic unit, embedded block RAM, rich routing resources, an underlying embedded functional unit, and an embedded dedicated hard core. Intel CEO Ke Zaiqi once described: "You can think of an FPGA as a bunch of gates that can be programmed at any time. FPGAs can be used as accelerators in multiple domains, such as facial search while encrypting, and in microseconds. Reprogramming. It costs far less than a large single custom component and has more flexibility."
With the development of technology and advances in technology, the performance and indicators of FPGAs have reached new heights. The new Arria10 family of FPGAs and SoCs consumes 40% less power than previous generation FPGAs and SoCs, and has the industry's only hard-core floating-point digital signal processing (DSP) module with speeds up to 1.5 trillion floating-point operations per second (1.5 TFLOPS) ).
Even with these advantages, the difficulty of developing FPGAs traditionally has made many programmers discouraged. However, with the advent of the OpenCL standard, the application threshold for FPGAs is being significantly reduced. Based on the OpenCL-based development framework, users can get rid of the traditional hardware description language (HDL)-based development process and switch to a more efficient software development process that is more in line with the programmer's background.
The OpenCL standard is the first open, royalty-free, unified programming model that accelerates algorithm implementation on heterogeneous systems and supports cross-platform development using C-based programming languages ​​such as CPU, GPU, DSP, and FPGA. Heterogeneous systems that support OpenCL typically communicate between the host and the hardware accelerator (or one core and another kernel without a host control) via the PCle bus, and this does not require the attention of the programmer. Instead, developers only need to understand the standardized application programming interfaces (APIs) defined by OpenCL. Today, all major deep learning frameworks provide OpenCL support. Intel is planning to introduce multiple network implementations of FPGA-based CNN.
The focus of our robotics lab this year is to build a heterogeneous computing platform based on CPU+FPGA for robots. Based on our experience, FPGAs can be of great value in some key robotic applications. For example, we ported the robot's own positioning algorithm (SLAM) to the FPGA for implementation, achieving a positioning frequency that is more than ten times higher than the CPU, allowing the robot to move at double speed without "disorienting."
In order to build a universal robot computing platform, we chose the CPU that Intel provides for the mobile platform and the Arria 10 series FPGA. Both communicate over high bandwidth, low latency I/O interfaces such as PCIe. A variety of acceleration modules can be included in the FPGA. We will provide a common interface for module developers to package and publish existing designs.
At the software level, we will build the configuration, installation, and messaging mechanisms for the acceleration module, so that third parties can integrate the required acceleration modules into the application and design common modules for positioning, control, visual recognition, and object modeling. Software integration package for users to use as a reference design or direct selection.
With such a wise brain, the robot will become more clever, with bright eyes, fast legs, steady hands, and integrate into our daily lives to provide better services for human beings.
There are a few different types of slip rings, but the most common type is called a cap slip ring. Cap slip rings have a cylindrical housing with an internal flange that fits over the end of the shaft. They typically have 6 or 12 channels and are available in a wide range of sizes.
Cap slip rings are used to provide electrical power and signals to and from rotating equipment. The channels on the slip ring allow for the passage of current and/or data signals through the ring. This allows for the rotation of devices such as antennae, radar dishes, and wind turbines without having to interrupt or disconnect the power or signal lines.
Cap slip rings are very reliable and can handle high speeds and heavy loads. They are also easy to install and maintain.
In the modern world, companies are always looking for ways to improve the efficiency of their machines and operations. Oubaibo offers a variety of products that can help improve your machine's performance. Their Cap Slip Ring allows for high-speed rotary unions, while their high-pressure swivel joints can handle even the most strenuous industrial applications. With so many options available, there's sure to be an Oubaibo product that can improve your machine's performance.
Cap Slip Ring,High Speed Rotary Unions,High Pressure Swivel Joints,High Pressure Swivel Joint
Dongguan Oubaibo Technology Co., Ltd. , https://www.sliprobo.com