Hubble adds Bluetooth event reporting to a Renesas edge AI project
Illustrative image: Unsplash contributor on Pexels
Hubble Network has integrated its connectivity SDK into the Edge AI Motion Tracker reference project in Renesas 365, combining a Renesas RA microcontroller and DA14531 Bluetooth Low Energy chip. The example is available now, with a hands-on workshop planned at electronica on 11 November.[1]
The RA device runs a Reality AI model that classifies motion from a sensor, recognising events such as a drop or circular movement. The Bluetooth device broadcasts the classification as a few bytes, allowing the remote application to receive the recognised event instead of a continuous stream of raw sensor samples.[1]
The accelerometer-based tracker also recognises shakes, giving a remote system information about how an asset has moved or been handled. Processing the motion locally keeps the recognition task on the device; the radio connection is then used to report a result that an asset management system can act on. The demonstration is a reference project, with the event model and application integration forming part of the development work.[2]
On Hubble’s terrestrial network, a nearby scanning gateway receives a Bluetooth advertisement and appends its own location, signal strength and location-accuracy metadata before forwarding the record. The backend produces a timestamped detection event, available to the application through an API or webhook. The location therefore comes from where a gateway saw the tag, rather than an implied GPS receiver in the tag itself.[3]
The time between an event being recognised and its arrival in the remote system depends on that radio capture. Hubble describes terrestrial detection as probabilistic, affected by the scanners present, signal conditions and the overlap between advertising and scanning schedules. Two devices travelling together can produce different packet counts because Bluetooth shares spectrum and mobile gateways may scan intermittently.[3][4]
Hubble’s guidance also identifies trade-offs in transmit power, beacon frequency and enclosure placement. More frequent advertisements give an intermittent scanner more opportunities to catch a message, while metal housings can attenuate the signal. These conditions affect reporting reliability even when the local motion classification has already completed successfully, separating a working AI demonstration from the performance of a deployed tracking service.[4]
The Munich workshop will take participants through the tracker on development hardware and show recognised events arriving on a dashboard, connecting the sensing, classification and reporting stages in one example.[1][2]
Abstract blue light trails illustrating wireless connectivity. Photo: Marek Piwnicki / Pexels. Image source. Free Pexels licence.
