Page 36

EETE APRIL 2013

ELECTRONICS IN ROBOTICS Robotics moves to multicore devices to create artificial nervous systems By Nick Flaherty A young German company is changing the way industrial robots are designed and developed. Some believe robotics today are at a point where personal computers were in the late 1970s, and sales figures bear this out, creating a huge opportunity for the industry to change says Synapticon in Baden-Württemberg. “We develop cyberphysical systems and one of the most exciting applications is service robotics,” said Synapticon’s managing director Nikolai Ensslen. “Tens of millions of dollars are invested in start-ups developing sensing robotics and this will be a $50bn market by 2025.” These ‘cyberphysical’ systems depend to a large extent on algorithms with a strong demand for computing power. While applications become more and more complex and need to run multiple compute-intensive tasks at the same time, the demand for computing power is rising continually but embedded computers using classical fixed instruction set processors usually whether do not provide enough computing power to run these applications at a reasonable speed or consume way too much energy for mobile use. Developers and manufacturers of robotic products today have to rely on components that are optimized for non-robotic purposes, spending time with overhead tasks rather than being able to concentrate on the development of ground breaking features. Moreover, mobile computing performance today still limits applications and this won’t change as long as PC-like computers are the workhorses, no matter how fast they are. “Today developers tend to use a couple of PCs running various operating systems that are not suitable for real time operations coupled with technology from the automation industry such as actuators and then have to control the hardware with microcontrollers and FPGAs. This is a very complex undertaking and a huge effort goes into developing the infrastructure,” said Ensslen. Instead Synapticon is using technology such as Core from UK-based startup XMOS Semiconductor to make use of a multicore architecture with guaranteed latency. This is a key advantage for developing industrial robotics, says Ensslen. Customers operating in robotics need to be able to integrate multiple motors (with sensors) in scalable, timing-critical electromechanical applications. Existing motor and motion control solutions are usually constrained to single functions and are not freely programmable, reducing the capabilities available. Synapticon’s core C22 module processor board uses xCORE to solve this problem: The compact 30x30mm board, based on two xCORE L16 chips, has the equivalent compute of a 32-core processor at a very small size and low power consumption. This allows customers to simultaneously control multiple motors over links up to 20m which is determined by the fixed latency. This gives substantially more flexibility in the system design, allowing sensors to be more widely distributed. Control loop computation takes place immediately at where sensor data is collected and motors are driven, leaving behind network latency issues and creating true artificial peripheral nervous systems. The technology has already been used by robotics companies such as Kuka for the mobile base of industrial robotics system. “Telling a biological arm to perform an action is done using an abstract command rather than giving each muscle separate multiple commands. Right now robotics generally works on this principle,” said Ensslen. “Using the Synapticon component based on XMOS we are able to achieve the same abstraction as we have in a biological being, enhancing the capabilities of robotics and providing a true differentiator and advantage in the market for our customers.” The deterministic nature of xCORE multicore microcontrollers is particularly useful in enabling Synapticon to implement both communications and control functions in a single platform, a task that would normally require multiple devices. In these applications XMOS can replace multiple devices with a single xCORE multicore microcontroller while a traditional solution a central controller would be required to communicate with each motor of a robotic arm separately. Fig. 1: The Kuka robotics base using Synapticon’s hardware and middleware running on XMOS multicore microcontrollers. Side-view and top-view. 36 Electronic Engineering Times Europe April 2013 www.electronics-eetimes.com


EETE APRIL 2013
To see the actual publication please follow the link above