The invention of battery-free Bluetooth® tags is inspiring the design of new solutions and a change in the architectural models used to build them. Solution architects and designers are developing ways of looking at how to structure battery-free Bluetooth solutions relative to existing ways of doing things. This Wiliot Solution Architecture Note examines the implications of integrating ultra-thin Bluetooth devices into IoT solutions and how this impacts the operating systems used, connectivity, edge processing, sensing, analytics, privacy, and security.
What are ultra-thin edges?
Traditional IoT devices are used as sensors collecting and sending telemetry to some remote service, to be stored and analyzed for various needs: predictive maintenance, security applications and many more. As we go left on the device spectrum there are less components, reducing the cost of the devices. On the far left, are the most minimal, ultra-thin devices composed from a single chip, glued on a polymer or paper inlay. There is no power source available on the device itself – all energy should be harvested from the environment.
The main constraint in ultra-thin edges is energy. Everything from transmission, data size, latency, processing, and sensing is governed by the ambient energy availability.
- Operating systems and configuration
Some of the edges have a dedicated operating system that can accept configuration and even version updates over a secured connection, for example, RTOS. Ultra-thin devices have a minimal embedded operating system and configuration, and due to its low cost and disposable nature it is very unlikely to have updates. In most RFID tags, there is no option for over-the-air (OTA) updates.
Unlike the devices we are familiar with today, connectivity of BLE and RFID devices is opportunistic. It depends both on the availability of nearby energy sources and data receiving gateways. In the world of RFID, it is usually the same gateway or reader that supplies energy to the tag and receives its transmissions, thus making it a synchronous ‘request-response’ like operation. Practically the gateway, acts as both energizer and gateway and data is expected to be available on sending energy on well-defined time slots.
Device-Gateway coupling. Most IoT devices are coupled to a specific gateway/client that transmits data from the device and receives commands and configuration from a remote service. In energy harvesting BLE tags, the nature of the operation is much more opportunistic. Data is sent from the devices upon the availability of an energy source nearby, which can be a mobile phone, Wi-Fi router or any other radio energy that can be harvested by the tag; however, nothing guarantees that a gateway will be in proximity to actually receive the data and transmit it to some remote service. In the case of MQTT, most current IoT frameworks rely on one-to-one relationships between a device and an MQTT client (and certificate), which makes the management of ultra-thin devices a challenge.
Long living connections. Long living connections like MQTT and Web Sockets are less convenient due to the opportunistic nature of gateway availability and the lack of a fixed coupling between the device and a gateway. Choosing transient connections like HTTP/S should be more reasonable in BLE beacons. However, since ultra-thin devices can send very small data packets (less than 100 Bytes), connection overhead, especially in HTTPS increase the actual data exchange between the gateway and remote services ending up with additional data and energy (for battery operated gateways) costs.
Multiplexing many devices on the same gateway connection can reduce data overhead. For example, by using HTTP 2, we can send multiple requests from many nearby devices using the same TCP connection and reduce both data and energy costs.
Data processing on the edge is minimal to non-existent on ultra-thin devices. Energy constraints are limiting both processing and state preservation and most of the data is analyzed on the gateway or a remote service. The opportunistic nature of data reception makes processing ‘energy expensive’ comparing to investing the energy in sending more transmissions, increasing the probability of data being received.
Sensing & Analytics
Since the expectation for the price of ultra-thin devices is significantly less than a dollar, sensors should remain simple and cost effective. Accurate, industry grade sensors remain out of scope and other more inaccurate sensors should be integrated into the devices on a very large scale. Combining these constraints with the volatile nature of the devices when ambient energy is not present, raises challenges that are alien to always-connected, always-sensing devices.
To overcome these challenges, analytics on the remote service is essential. The first requirement is to overcome sensor variability and inaccuracies by building models to calibrate sensing data in real-time. Calibration is done both for a single device and device population, based on the physical properties of the sensed metric (pressure, temperature etc.) and the location of nearby devices. In addition to calibrating raw telemetry, synthesis of derived events such as the “drop” or “pickup” of a product/package with an ultra-thin sensor inside, yields high value data for specific product analytics such as last mile delivery, sleep analytics and many more.
Privacy & Security
Ultra-thin devices are usually present in consumer environments, embedded in things such as: garments, furniture, home appliances, tickets and store shelves. This requires protecting people’s privacy and avoiding unauthorized tracking of locations, activities and other sensitive information. Protecting privacy should be enforced through the entire data chain: device > gateway > remote service and be embedded into the holistic design of the system.
These requirements, combined with the opportunistic nature of nearby gateways creates challenges, for example, how to avoid device tracking. Today, even without understanding the content of transmitted data, one can track Bluetooth Low Energy advertisement packets and link them to a specific device. In order to avoid that, the entire data transmitted should be encrypted and anonymized in a way so that only the data or device owner can detect the device identity and understand the transmitted data. The most secure way to achieve that is by using a remote service that handles identity resolution and data decryption, thus avoiding unauthorized sniffing and gateway vulnerabilities.
Creating IoT solutions that incorporate ultra-thin edge devices requires special considerations in system architecture as demonstrated by Wiliot’s solution. These low cost devices enable architects, developers, and data scientists to build new, exciting solutions at unprecedented scale for both consumer and industrial applications.