The increasing importance of Edge Computing in Smart City applications
Until very recently, workload in Smart City applications has generally been managed in the cloud, even if that implied the transmission of huge data volume (for example, video streams) and adverse latencies for certain use cases.
Recent advances in processing capacity, power management, and wireless communications (from LPWAN to 5G), however, are enabling embedded systems to assume part of the processing load of Smart City applications on the edge or even endpoints, which means, nearer to the data sources and reducing notably response times.
A very interesting example is the partnership collaboration between NVIDIA and Nota AI in the deployment of Smart Traffic Solutions, such as real-time traffic lights control to speed up emergency transportation, using image recognition and Reinforcement Learning models
The combination of Edge Computing and Cloud Computing results in hybrid systems; the end points send data to edge servers, where a first stage analysis is performed and, using a rules engine, the information is transferred to other servers on the edge and/or the cloud.
For instance, if the sensor in a smart public trash bin detects smoke, the street cleaning operator in the zone is alerted immediately through edge systems, and the information is also up scaled to cloud systems. The inherent complexity of these hybrid systems is even greater if we consider the requirement of deploying software updates to a huge number of devices that, worse yet, can be physically very different. Fortunately, Container Orchestration resolves comprehensively this problem, even with open-source solutions for servers, like Red Hat Openshift, and others more oriented to fleet management, like OpenBalena.
A new generation of very low power consumption (also very low cost) microcontrollers, able of running AI/ML models using batteries or energy harvesting, is currently pushing the border of Edge Computing towards the endpoint, even in IoT and wearable applications. Some companies, like Sony with its Spresense microcontroller are betting on multicore solutions, while others, like Syntiant Corp. prefer to develop neural network accelerators.
This increasing trend towards transferring workload to the end point is reaching the physical limits of miniaturization with smart sensors like Bosch Sensortec GmbH BHI260AP that in less than 5x5 mm includes a microcontroller that can run self-learning functions and an IMU (Inertial Measurement Unit) enabling dead reckoning even without GPS assistance.
All these technologies converge in Smart City solutions and require technicians and engineers aware of its possibilities and its limitations. Talented experts will envision the future of Smart Cities through the continuous attainment of brand-new knowledge and skills.
