The Internet of Things (IoT) is expanding and powering smart homes with a wide range of connected devices like smart appliances (lights, ovens, coffee makers, refrigerators, etc.), security systems (surveillance cameras, video doorbells, locks, etc.), and home entertainment systems (TV, music systems, voice assistants, etc.).
In the current IoT paradigm, these connected devices are constantly gathering huge amounts of data and sending it to a centralized remote server. This data is processed and stored in the remote server for analytics and remote access control. Centralized remote servers have the power to process this data, however, there are multiple limitations in this architecture.
Here are the top three IoT data challenges:
- Sending the huge data generated from connected devices to the remote server causes latency issues and delays in the transfer of data. This is a major issue specifically in real-time analytics and time-critical actions
- The high-security risk of the data in flight from smart home to remote server and the data stored on the remote server
- Privacy concern with complete information of smart home going outside & getting stored in a distant server
“Annual global cloud IP traffic will reach 19.5B by the end of 2012.”
– Cisco Global Cloud Index
With the advent of edge analytics and fog computing, the above challenges are addressed, bringing intelligence, performance, security, and privacy in smart homes:
- Edge Intelligence is based on the concept to bring data processing from the cloud to the field, i.e. to the smart home where sensors, devices are deployed.
- Depending on the processing requirement, the data collected from the device is either processed at the device itself known as edge analytics or at the local node deployed at the periphery of the home network known as fog computing.
- The analysis and decision taken at edge or fog node are fed to the device for the next action in real-time.
- Only a subset of data is transmitted to a remote server. Sensitive information can be processed at Edge allowing only non-sensitive information to be sent to the server. This helps in real-time data processing without latency, reduces privacy/security risks, and optimizes the network resources.
Figure 1: Edge Computing Architecture
Power of Edge Intelligence in Super Smart Home
Here are some smart home scenarios to see how edge intelligence will help in faster, secured real-time processing without the need of being connected with the remote server:
Security: Smart security systems including security cameras, video doorbells, and alarms are sending data to fog nodes deployed in the home network. This data is processed locally on the fog node to detect unsolicited activity, sending notifications, and raising alarms. This helps avoid a security system being compromised at the 3rd party server.
Privacy: Security cameras processing the images and voice recognition to detect unsolicited activity at the edge. This helps in ensuring that private and sensitive data from streams of camera or microphones do not leave premises without consent.
Internet Connectivity, Bandwidth & Latency: Smart video locks processing of the images and voice to open the door automatically even in-network outage and avoiding network latency.
Edge Analytics and Fog Computing: Fitness bands monitoring and analyzing the user’s vital parameters and prescribing the precautionary measures or even calling an ambulance in case of emergency.
Fog node aggregating the data collected from the fitness band and indoor cameras, analyzing this data to detect user mood. Fog node is then customizing the home environment according to the user mood e.g. on being asked to play the songs as per user mood, smart speakers play the relevant playlist, etc.
Safety: Fire safety by triggering a call to fire emergency service in case of fire.
Troubleshooting of home network devices and recommending user to fix network issues.
There are plenty of other real-time scenarios specifically in low connectivity areas, where edge analytics and fog computing can be a boon to enhance user experience.
Competitive Landscape for Edge Intelligence
Edge intelligence is triggering the interest of both IoT platform service providers and smart device manufacturers to create innovative solutions.
IoT Platform Service Providers
Most of the major platform service providers that provide IoT cloud solutions have their own edge computing solutions. These edge solutions are integrated with cloud solutions to provide seamless connectivity and deployment. However, interoperability is still a challenge where the edge computing solution of one platform service provider cannot be directly integrated with the cloud solution of another service provider. Below are the edge computing solutions provided by leading IoT platform service providers:
AWS Green grass:
- Enables the local execution of AWS Lambda, messaging, data caching, and security
- Devices that run Linux and support ARM or x86 architectures can host the AWS IoT Greengrass Core. These devices can vary in size, from smaller micro-controller-based devices to large appliances
- Edge devices operate reliably and securely even when they are offline. Once the device reconnects, it synchronizes the data on the device with AWS IoT Core
Google Cloud IoT Edge
- Cloud IoT Edge is composed of two runtime components – Edge Connect and Edge ML. It also takes advantage of Google’s purpose-built hardware accelerator ASIC chip, Edge TPU
- Edge connect is used to securely connect edge devices to the cloud, enabling software and firmware updates, and manage the exchange of data with Cloud IoT Core
- Edge ML is used to run ML inferences of pre-trained TensorFlow Lite models locally
- Edge TPU augments Google’s Cloud TPU and Cloud IoT to provide an end-to-end (cloud-to-edge, hardware + software) infrastructure to facilitate the deployment of customers’ AI-based solutions
Smart Device Manufactures
Below are some smart device manufactures that are taking steps forward in bringing intelligence to the device:
- Nest Cam IQ security camera uses Google AI for image recognition and video processing to generate person alerts. Processing the data within the device itself enables us to take instant action even before the data is sent to the cloud.
- Butterfleye or Ooma Smart Cam cameras offer artificial intelligence at the device to recognize faces, people, and many other objects. It continues to work during internet and power outages with built-in storage and battery backup
There are other device manufacturers also working in the direction to move analytics at the edge on the device.
What’s next for the Smart Home?
To take the next leap in IoT, smart homes need to be upgraded to intelligent homes that leverage the capabilities of artificial intelligence (AI) and machine learning (ML). Current IoT architectures depend on centralized servers to do the processing of the data collected from connected devices. This adds security and privacy risk along with the latency in time-critical actions. Edge intelligence is an evolving technology to solve these challenges of traditional IoT cloud architecture. Edge intelligence moves the analytics within the periphery of the home network to sustain the volume and velocity of the data without compromising the privacy and security of the data collected.
Home automation leaders should consider that not all the applications or processes are suited for either remote server or edge. The smart home ecosystem should be analyzed to optimize the deployment with a mix of the remote server and edge solutions in order to transform a smart home to a truly intelligent home.
With strong expertise in IoT and machine learning, we at Capgemini Engineering are here to help OEMs and IoT solution providers build optimized edge computing ecosystems to lead them to the future.
- Left Shift Cloud Intelligence to the Edge
- How Edge Computing Unlocks the Networks of the Future
- Taking Compute Power to the Edge
Arpna Gupta Senior Engineering Project Manager
MEET OUR EXPERTS
Contact a Capgemini Engineering edge intelligence expert today to find out how smart homes can truly become Intelligent.