05 Aug 2016
Increasing solution differentiation through edge-based heterogeneous computing
By any measurement or however one defines it, the Internet of Things (IOT) is big, and is getting even bigger! Numerous companies are looking to design IOT embedded devices or devices that have communication capabilities, sending data to other devices or to cloud services. From automotive to wearables, home automation to drones, IOT encompasses many different markets. Each one of these markets also has plenty of competition (how many fitness trackers exist today?) The embedded device companies that can truly differentiate their products from their competition will have the highest likelihood of success. Ask yourself, how do you differentiate your product offering?
Source: Internet of Things: Are We There Yet? (The 2016 IoT Landscape) Matt Truck http://mattturck.com/2016/03/28/2016-iot-landscape/
Many IOT devices on the market today have been simple in nature, collecting limited sensor data and sending data upstream. Unfortunately, with very limited processing capabilities, these remote devices have very limited security—if any, limited device interaction capabilities, and very low rate data path. The solutions are attempting differentiation via cloud services, which is proving that true device differentiation has remained somewhat elusive. Evaluations of the industry have demonstrated major obstacles for IOT adoption include security, remote device management, lack of standards for inter-device communication, upgradeability of the solutions, power consumption and adopting the right business models (ongoing service pricing) that customers are willing to pay for, among others.
Adding more capable processing at the edge will enable true solution differentiation, allowing flexibility for more intelligent capabilities at the point of data collection. Developers can leverage the latest advances in sensors and computing, some of which are developed for the mobile market, in addition to other advancing capabilities such as data analytics, artificial intelligence, machine learning, data mining, and machine vision, in order to differentiate their solutions. Combining many different types of sensors (video, sound/acoustic, etc) and modern diverse sensor technologies in a power-efficient solution will also provide more context and enhance situational awareness in the physical world, thus providing more value for your customers.
Source: Sensor processing done right: Smart integration of a specialized sensor engine | Qualcomm https://www.qualcomm.com/news/onq/2014/06/05/sensor-processing-done-right-smart-integration-specialized-sensor-engine
The most power efficient way to access the performance necessary for this differentiation on an embedded device is with heterogeneous processing, where you can run each specific computing task with a purpose built, optimized processing core. For example, while a CPU can be used for general system tasks like keeping track of the system state or managing files/general operating system tasks, a DSP can be used for filtering incoming data from multiple sensors in real time while using very limited power. Or a GPU can be used to process large parallel arrays of data very efficiently, such as image analysis, data mining or encryption/decryption. These specialized tasks running on purpose-built processors lowers the total amount of power necessary to perform the tasks. Add it all up, you have very high performance at a fraction of the power budget.
Instead of edge devices sending limited sensor data upstream allowing little to no interaction, consider adding more processing capabilities at the edge to truly differentiate your solution. Set your solution apart with more efficient, higher performance processing analyzing local sensor data in real time, limiting the amount of data necessary to communicate while allowing more complex interaction with other devices/cloud, increasing security, flexibility, upgradeability and convergence of the entire solution.
Interested in exploring how Intrinsyc can help you design, develop, and/or build your next product based on power-efficient heterogenous processing? Please email our sales department firstname.lastname@example.org
Steve Bath is a Solutions Architect for Intrinsyc Technologies. With 16 years of experience in the semiconductor, IP video surveillance and product development/design services markets, he has a unique perspective helping companies develop successful embedded products. Steve has a Bachelors of Science in Computer Engineering from the University of Notre Dame.