Decentralized computing power not only relieves the networks, it also shortens latency. Dr. Ahmad Bahai, CTO at Texas Instruments, expects that even AI applications will be running in the Edge in the future.
The spread of distributed autonomy in fields as diverse as industrial, automotive and medical is unprecedented. Increasing numbers of connected sensors provide an enormous amount of valuable data that can be used to improve the performance and efficiency of factories, cars and cities. However, due to concerns about latency, bandwidth and energy limitations, not every piece of data should or could be processed in the cloud.
The availability of intelligence at the edge of the network can significantly improve the overall performance of many applications. A combination of embedded connectivity, processors and power management provides a platform for embedded intelligence. An intelligent node often is the best option to process and analyze local data. But it must do so on a low power budget.
Low-power, agile, embedded radios are an indispensable part of intelligent sensors and transducers. Ultra-low power embedded processors with multiple I/O interfaces and peripherals, optimized memory (including in-memory computing), integrated clocking subsystems and neural network accelerators can run some real-time Artificial Intelligence algorithms quite efficiently. Creative power management, with a hybrid combination of batteries and energy harvesters, can monitor and manage applications with nano-power accuracy. In some applications, the availability of embedded MEMS sensors results in the further integration of intelligent nodes.
Innovations in semiconductor technology are enabling intelligence everywhere – all the way to the edge.