“Robots continue to revolutionize the future of work through increased efficiency by automating repeatable tasks and making complex decisions autonomously. The better a robot can adapt to changes in its environment, the more value it can provide.
By Susan Cheng, North American Industrial Vision Marketing Manager, Xilinx
As robots can adapt to changing environments, their value and potential impact are climbing rapidly.
Robots continue to revolutionize the future of work through increased efficiency by automating repeatable tasks and making complex decisions autonomously. The better a robot can adapt to changes in its environment, the more value it can provide.
The electronics industry has long relied on application-specific integrated circuits (ASICs) to provide the high performance and real-time responsiveness that robotics applications demand. However, with the demand for more capable robots and their ability to adapt to newer environments, robots must have the ability to dynamically optimize as the latest artificial intelligence (AI) models evolve. With the continuous evolution and development of the AI algorithms underlying these AI models, the hardware responsible for AI acceleration tasks must also have the ability to adjust accordingly at any time.
ASICs are inadequate not only because they typically require an 18-month development cycle, but also because any major change requires iterating the ASIC, executing the development process from scratch.
In short, for robots to be able to adapt to changing applications and needs, the technology that makes them needs to be equally flexible. Conversely, when we put an upper limit on the performance of the robot, we also attach a “limitation” or constraint to a specific technology. Their limited effect of curing is like a pizza robot that claims to be able to choose any menu, but can only make two flavors of cheese and pepperoni.
The problem is that predicting future developments, especially robotics and AI technology in general, is extremely difficult. As the capabilities of robots increase, so do the challenges they must overcome. At the same time, robots will need to change the way they move, see, think, and interact with the world, requiring the underlying technology platforms used to build robots to be flexible enough to support these new skills while providing real-time responsiveness. of advancing with the times.
Complicating this conundrum is the fact that even after the robot is deployed in the field, it needs to continue to adapt to changes in its environment. These include being able to optimize for new AI algorithms and the latest features.
The key to building a robot with long-term adaptive capability is to build a robot directly on a technology platform that is inherently flexible. For example, Xilinx offers robot manufacturers an innovative adaptive computing platform. Combining hardware-based acceleration performance with software-programmable flexibility, this platform is an ideal development platform for building flexible and adaptable robots.
Adaptive computing devices such as FPGAs and adaptive SoC devices are like chameleons. When the robot needs to change its algorithm, the adaptive computing architecture can dynamically update with it. Adaptive computing platforms can provide over-the-air (OTA) updates not only for software, but also for hardware. In this way, for any workload, the adaptive computing platform can be reconfigured to be the best processor for the latest tasks.
The significance of an adaptive computing platform is far more than providing future-proof compatibility for designs, but also the ability to meet multiple applications and multiple market needs by providing a single hardware design. This capability enables users to significantly extend the effective life of their designs. As a result, robot manufacturers can take advantage of the economies of scale of a single design over multiple sets of different designs, while saving development time and development costs.
By building robots based on adaptive platforms, manufacturers can also continue to introduce new capabilities into the system after the robots are in the field, and are expected to create new revenue streams. For example, robots will be able to track their day-to-day operations using state-of-the-art predictive maintenance algorithms and predict when failures will occur. This helps robots determine maintenance time before failure occurs, increasing reliability and reducing costly downtime.
Change is an immutable truth in the IT industry. The speed and complexity of field change in AI and robotics is much higher than in other market segments. Adaptive computing platforms can not only help robot manufacturers deliver advanced capabilities, but also ensure they have the agility to continue delivering advanced capabilities long after other vendors’ fixed-function ASIC designs have long since failed.
It’s a hard fact that robotic systems that can adapt to changing environments will replace those that don’t. Manufacturers also cannot escape this law. Manufacturers who can take advantage of the most advanced technologies with the fastest speed and greatest flexibility will win the greatest opportunity in the robotics market.