Smart Beacons Leverage Bluetooth System-on-a-Chip for Connected ML Insights

Today’s product development and support cycles move fast. Embedded products that detect software and hardware faults and provide insight into user behavior, supply the data that engineers need to keep equipment working and improving.

But not all industrial equipment is wired for easy connectivity to support these embedded products. Even products designed for the Internet of Things (IoT) can encounter connectivity issues such as electromagnetic interference (EMI), limited bandwidth, and long cable runs.

The emergence of Bluetooth-enabled System-on-a-Chip (SoC) technology gives engineers access to seamless connectivity alongside microprocessor power that supports on-board machine learning (ML). This pairing of connectivity with smart analytics is a valuable tool in a proactive, instead of reactive, design and support cycle.

Smart data collection transforms product development and support

Successful product development and support require usage data. Designers who don’t know how customers are using a product, including which features they rely on and which ones are cumbersome or buggy, will have trouble iterating the product to an upgrade users want. Likewise, support personnel can’t adequately troubleshoot problems without knowing user behavior, system status, environmental conditions, and other key data right before or during the problem.

A product with modern onboard connectivity and analytics can make both design iteration and support more effective. Embedded products and smart beacons can detect environmental conditions like temperature, humidity, and barometric pressure, as well as sensing acceleration in multiple axes, ambient light, and magnetic fields. Timestamps from a real-time clock (RTC) allow the data to be correlated with other system events, either using onboard analytics or when broadcast to a cloud server via Bluetooth.

For example, a smart beacon attached to a linear motion system in an industrial environment might detect that vibrations increase when humidity is elevated. On-board processors could then broadcast an alert to maintenance engineers that additional lubrication is needed. This kind of proactive troubleshooting reduces equipment downtime and maintenance costs.

Product designers might also use the logged vibration and environmental data to improve future versions of the linear motion system. For instance, they might recommend a different lubricant that lasts longer in humid conditions. They might also redesign the lubrication system to better protect it from the elements.

Implementation challenges and solutions

In order to realize the benefits of enhanced data collection in an IoT setting, engineers must optimize data collection and analysis. Any transfer of information to the cloud for analysis has inherent latency and reduces data security. Embedded systems and smart beacons combat this by packaging AI and ML capabilities into the units themselves. These edge AI and TinyML systems contain scaled-down software models that allow the processors to make intelligent inferences based on the real-world data they receive.

Onboard ML capabilities can be as simple as matching vibration data, environmental data, and a global timestamp, or complex enough to predict maintenance needs based on data trends. Whether complex or straightforward, the ML module receives and processes real-time data without utilizing network resources, resulting in timely insights and minimal energy use.

Still, smart beacons and embedded systems eventually need to communicate status to other devices or to a server via a network. Many legacy systems are designed for wired serial connectivity with protocols like PROFIBUS, DeviceNet, CANOpen, and Modbus RTU. More modern equipment relies on low-latency Ethernet-based protocols like PROFINET, EtherCAT, EtherNet/IP, or Ethernet POWERLINK. However, both serial and Ethernet communication require cables for data and power across factory floors with the attendant challenges of EMI, signal degradation over long cable runs, and the facilities investment needed to mitigate trip hazards and provide pathways for driven or autonomous vehicles.

Short-range, radio-frequency (RF) communication using Bluetooth protocols overcomes many of these challenges. Some versions of Bluetooth, like Bluetooth Low Energy (BLE), are designed to broadcast strong signals up to 150 m on the power available from a button battery, removing the need for both power and data cables.

A BLE signal runs in the 2.4 GHz band that also supports some cellular and Wi-Fi networks. While the shared band can lead to network interference and reduced signal integrity, it is also the most reliable band for overcoming line-of-sight obstacles, such as walls and equipment. To overcome line-of-sight and interference concerns, many BLE systems can be mesh networked to utilize Internet Protocol version 6 (IPv6) to connect BLE devices to each other and to the cloud (Figure 1). Strategically placed Bluetooth hotspots can also boost signal strength and integrity within the mesh network.

Figure 1: Smart beacons and other devices can use Bluetooth to connect to the nearest hotspot without pairing. Hotspots can enable Bluetooth mesh networks or connect to cloud services via Wi-Fi. (Image source: Blecon LTD)

Smart beacons bring analytics and networking together

Marrying data collection, AI and ML inference engines, and network connectivity, Bluetooth-enabled smart beacons provide product operation, user behavior, and predictive maintenance insights, even on equipment that’s not designed for embedded systems. One example is the L02S-BCN by Blecon LTD (Figure 2).

Figure 2: L02S-BCN smart beacons have BLE connectivity, multiple sensing options, high-visibility LEDs, and a field-replaceable battery in an IP67 enclosure. (Image source: Blecon LTD)

L02S-BCN smart beacons are driven by nRF54L15 series multiprotocol SoCs (Figure 3) from Nordic Semiconductor. These chips combine a multiprotocol 2.4 GHz radio that supports Bluetooth version 5.4, IEEE 802.15.3-2020, and 2.4 GHz protocols at data transmission speeds up to 4 Mbps with a 128 MHz Arm® Cortex®-M33 processor running on 265 KB of RAM. The 1.5 MB of non-volatile memory can store sensor readings and analyses if network connectivity is unavailable.

Figure 3: The nRF54L15 series multiprotocol SoCs have multifunction radios, PSA Level 3 security, a 128 MHz processor with 256 KB RAM, and hardware and software peripherals that support edge AI and ML. (Image source: Nordic Semiconductor)

The nRF54L15 chip has built-in security designed for IoT systems. TrustZone isolation, side-channel protection, and tamper-detection protocols certify it to platform security architecture (PSA) Level 3. These systems ensure data payloads transmitted from L02S-BCN beacons are encrypted for secure transport and that network node identities are verified from the cloud via two-way communication.

The nRF54L15 chips also have built-in peripherals that allow L02S-BCN smart beacons to collect, analyze, and share data from IoT systems. A 14-bit analog-to-digital converter (ADC) translates signals from temperature, humidity, barometric pressure, acceleration, and photosensitive sensors into digital data, while a global RTC creates a time stamp for each reading. Five serial interfaces, including serial peripheral interfaces (SPIs), two-wire interfaces (TWIs), and universal asynchronous receiver/transmitters (UARTs), connect processing and sensing components.

In addition to these physical sensor options, L02S-BCN beacons also act as embedded devices, using pre-integrated Memfault software on open standard Reduced Instruction Set Computing version five (RISC-V) coprocessors to detect and report crashes, software faults, battery status, and user behavior to the cloud. Memfault also manages over-the-air (OTA) updates, so there’s no need to recall deployed devices.

L02S-BCN beacons also demonstrate the use of Edge Impulse, an edge AI platform, to deliver ML without using network resources. Edge AI removes latency and allows L02S-BCN beacons to operate on 1000 mAh CR2477 button batteries that can be replaced in the field. The 69.9 mm tall by 46.7 mm wide by 18 mm thick L02S-BCN beacons have IP67-rated enclosures that exclude dust and withstand immersion in 1 m of water for up to 30 minutes. The beacons can be mounted to equipment using double-sided adhesive, screws, or zip ties.

Conclusion

Bluetooth smart beacons bring sensing, connectivity, and edge AI and ML to industrial and IoT applications. Powered by SoCs like Nordic Semiconductor’s nRF54L15 that support data collection, zero-latency analytics, and OTA updates, smart beacons such as Blecon’s L02S-BCN overcome barriers to connectivity to turn industry-deployed equipment into embedded products with ML capabilities.