Category: AI

TinyML and Bluetooth

There’s interesting new research into TinyML for On-Device and Edge Analytics in Wireless Networks. The beacon-relevant parts are concentrated around BLE indoor navigation, proximity detection, RF fingerprinting-based positioning and beacon transmission-power control.

The paper describes RF fingerprinting-based indoor positioning, where radio infrastructure such as Wi-Fi access points, BLE beacons and cellular base stations can be part of the positioning system. It explains that fingerprints are collected in an offline phase, used to train a positioning algorithm, and then used online to infer an asset’s location. The paper says tiny deep learning is attractive here because it can extract features automatically, handle non-linear boundaries in indoor spaces, and run at the edge with lower latency, better privacy and less reliance on a central server. It cites an indoor asset-tracking implementation using BLE beacons that achieved about 88% location-estimation accuracy.

For BLE navigation, the paper treats beacons as especially relevant in indoor places where GPS is unreliable, such as hospitals, shopping centres, factories and university buildings. In that setting, mobile devices estimate their position using RSSI measurements from nearby BLE beacons; RSSI is the received signal power level and is also used to infer proximity between devices.

The main proposed tinyML opportunity is RSSI-based transmission-power management on BLE beacons. The idea is that beacons send periodic messages to mobile user devices, while also using RSSI from other beacons’ transmissions to estimate interference. A tiny reinforcement-learning model could run directly on each beacon, using received signals as the state, transmission power as the action, and a reward based on the resulting RSSI. The purpose is to tune beacon power continuously, reduce interference and improve indoor navigation performance.

The other beacon-adjacent opportunity is improving proximity detection through RSSI-based ranging. The paper argues that classical RSSI ranging is weak because it mostly relies on spatial features and ignores time- and frequency-domain RSSI fluctuations; motion between devices adds fading, reducing confidence and accuracy. It suggests using tiny deep learning on a mobile MCU-based receiver to incorporate broader RSSI features, including time-domain, frequency-domain and statistical characteristics, with the aim of improving distance estimates.

In practical terms, the paper’s message is that BLE beacons are a good fit for tinyML because the intelligence can move closer to the beacon or mobile device.

New Bluetooth Low Energy Error Correction Using AI

A new paper Error Correction in Bluetooth Low Energy via Neural Network with Reject Option by Almeida et al. (2025) presents a new method for improving data reliability in Bluetooth Low Energy (BLE) communication without modifying the transmitter. The technique combines cyclic redundancy check (CRC) error detection with a neural network that has a reject option, allowing it to identify and correct bit errors more effectively.

The study explains how BLE devices, particularly in Internet of Things (IoT) applications, suffer from data corruption due to multipath fading and interference. Traditional error-correcting codes, such as Turbo or LDPC, are unsuitable for BLE because of their computational and memory demands. Instead, the authors propose an Extreme Learning Machine (ELM) neural network that detects uncertain bits using a reject output (labelled R) and then flips them for CRC revalidation, iterating until the packet is corrected.


Simulations using Rayleigh fading and additive white Gaussian noise channels showed that the method achieved correction rates between 94–98% for single-bit errors and 54–68% for double-bit errors, depending on packet size. It significantly lowered packet error rates and improved throughput compared with uncorrected transmission.

When applied to compressed grayscale image transmission, the method restored visual quality under noisy conditions (signal-to-noise ratios of 9–11 dB). Measured using Structural Similarity Index (SSIM) and Peak Signal-to-Noise Ratio (PSNR), image quality improved markedly, often recovering most of the lost detail.

The approach outperformed other CRC-based correction algorithms such as CRC-ADMM and CRC-BP while requiring less computational power and memory. Processing times were substantially lower, enabling near real-time correction suitable for BLE and IoT devices.

The proposed neural network with reject option offers an efficient, scalable, and energy-aware method for enhancing BLE reliability without additional transmitter complexity. It reduces retransmissions, improves data integrity, and enhances performance in both data and multimedia transmission scenarios.

Improved RSSI Indoor Localisation Using AI Algorithms

The article titled Improved RSSI Indoor Localization in IoT Systems with Machine Learning Algorithms by Ruvan Abeysekera and Ruvan Abeysekera focuses on enhancing indoor localisation in Internet of Things (IoT) systems using AI machine learning algorithms. The paper addresses the limitations of GPS in indoor environments and explores the use of Bluetooth low-energy (BLE) nodes and Received Signal Strength Indicator (RSSI) values for more accurate localisation.

GPS is ineffective indoors so the paper emphasises the need for alternative methods for indoor localisation, which is crucial for various applications like smart cities, transportation and emergency services.

The study uses machine learning algorithms to process RSSI data collected from Bluetooth nodes in complex indoor environments. Algorithms like K-Nearest Neighbors (KNN), Support Vector Machine (SVM, and Feed Forward Neural Networks (FFNN) are used, achieving accuracies of approximately 85%, 84%, and 76% respectively.

The RSSI data is also processed using techniques like weighted least-squares method and moving average filters. The paper also discusses the importance of hyperparameter tuning in improving the performance of the machine learning models.

The research claims to provide significant advancement in indoor localisation, highlighting the potential of machine learning in overcoming the limitations of traditional GPS-based systems in indoor environments.

An AI Machine Learning Beacon-Based Indoor Location System

There’s a recent paper by researchers at DeustoTech Institute of Technology, Bilbao, Spain and Department of Engineering for Innovation, University of Salento, Lecce, Italy on Behavior Modeling for a Beacon-Based Indoor Location System.

The research compares two different approaches to track a person indoors using Bluetooth LE technology with a smartphone and a smartwatch used as monitoring devices.

The beacons were iB005N supplied by us and it’s the first time we have been referenced in a research paper.

The research is novel in that it uses AI machine learning to attempt location prediction.

The researchers were able to predict the user’s next location with 67% accuracy.

Location prediction has some interesting and useful applications. For example, you might stop a vulnerable person going outside a defined area or in an industrial setting stop a worker going into a dangerous area.

Positioning, Tracking and Flow Using Bluetooth Signals

There’s new research into Developing a Solution for Mobility and Distribution Analysis Based on Bluetooth and Artificial Intelligence.

The paper describes an efficient solution for locating, tracking, analysing distribution and flow of people and/or vehicles. Filters and algorithms including artificial intelligence and angle of arrival (AoA) were employed.

Locating using Triangulation
Triangulation Using Angle of Arrival (AoA)

The resultant system provided for analysis of location, traffic flow and passenger movement along routes.

3D RSSI Variation with distance

The researchers found that accuracy was improved when multiple measuring stations were used. Improved positioning was achieved using geometry algorithms (Voronoi) and the k-mean cluster algorithms. Artificial intelligence allowed for deeper analysis of the data for more accurate positioning, trajectory estimation and density evaluation.

The State of AI in 2019

Beacons provide a great way of providing new data for AI machine learning. They allow you to measure things that aren’t currently being quantified, create new data that isn’t silo’d by protectionist staff or departments and allow you to pre-process data in-place making it suitable for learning and inference.

There’s a new free State of AI Report 2019 in the form of a 136 page presentation. It covers aspects such as research, talent, industry and geopolitical areas such as China and Politics.

Read more about AI Machine Learning with Beacons

The Crux of Machine Learning is Realistic Expectations

Venturebeat has an article, based on IDC research, titled For 1 in 4 companies, half of all AI projects fail.

“Firms blamed the cost of AI solutions, a lack of qualified workers, and biased data as the principal blockers impeding AI adoption internally. Respondents identified skills shortages and unrealistic expectations as the top two reasons for failure, in fact, with a full quarter reporting up to 50% failure rate.”

We believe a key part of this is ‘unrealistic expectations’. Half of all AI projects failing for 1 in 4 companies isn’t unreasonable. AI and machine learning should be viewed as a research rather than a development activity in that it’s often the case that it’s not known if the goal is achievable until you try.

Another unrealistic expectation of machine learning is often to have 100% accuracy. The use of an accuracy % in assessing machine learning models focuses stakeholders minds too much on the perceived need for a very high accuracy. In reality, human-assessed, non-machine learning, processes such as medical diagnosis tend to have much less than 100% accuracy and sometimes have undetermined accuracy but these are reasonably seen as being acceptable.

In summary, there has to be upfront realistic expectations of both the possible outcome and the accuracy of the outcome for projects to correctly determine if AI activities are an unexpected failure.

Read about AI Machine Learning with Beacons

Using AI Machine Learning on Bluetooth RSSI to Obtain Location

In our previous post on iBeacon Microlocation Accuracy we explained how distance can be inferred from the received signal strength indicator (RSSI). We also explained how techniques such as trilateration, calibration and angle of arrival (AoA) can be used to improve location accuracy.

There’s new research presented at The 17th Annual International Conference on Mobile Systems, Applications, and Services (MobiSys ’19) by researchers from Nagoya University, Japan that looks into the use of AI machine learning to process Bluetooth RSSI to obtain location.

Their study was based on a large-scale exhibition where they placed scanning devices:

They implemented a LSTM neural network and experimented with the number of layers:

They obtained best results with the simplest machine learning model with only 1 LSTM:

As is often the case with machine learning, more complex models over-learn on the training data such that they don’t work with new, subsequent data. Simple models are more generic and work not just with the training data but with new scenarios.

The researchers managed to achieve an accuracy of 2.44m at 75 percentile – whatever that means – we guess in 75% of the cases. 2.44m is ok and compares well to accuracies of about 1.5m within a shorter range confined space and 5m at the longer distances achieved using conventional methods. As with all machine learning, further parameter tuning usually improves the accuracy further but can take along time and effort. It’s our experience that using other types of RNN in conjunction with LSTM can also improve accuracy.

If you want to view the research paper you need to download all the papers from the conference (zip) and extract p558-uranoA.pdf. Some of the other papers also make interesting, if not directly relevant, reading.

Read about AI Machine Learning with Beacons

Free AI Paper

Microsoft has a new free (registration not required) paper on Maximising the AI opportunity, How to harness the potential of AI effectively and ethically (pdf). While the data is UK centric, the insights and actions are applicable to any country.

The message is that organisations should embrace AI’s potential or risk being left behind. As well as economic gains, changes should take into account social and safety issues.

“Organisations that are investing in establishing the right approach to AI now outperform those that don’t by 9%”

The paper explains AI and how many organisation are talking about AI but fewer are taking action. It gives perspectives of use of AI in FinTech, Healthcare, Manufacturing and Retail.

Read about AI Machine Learning with Beacons

Machine Learning isn’t Magic

When working with Machine Learning on beacon sensor data or indeed any data, it’s important to realise AI machine learning isn’t magic. It isn’t foolproof and is ultimately only as good as the data passed in. Because it’s called AI and machine learning, people often expect 100% accuracy when this often isn’t possible.

By way of a simple example, take a look at the recent tweet by Max Woolf where he shows a video depicting the results of the Google cloud vision API when asked to identify an ambiguous rotating image that looks like a duck and rabbit:

There are times when it thinks the image is a duck, other times a rabbit and other times when it doesn’t identify either. Had the original learning data included only ducks but no rabbits there would have been different results. Had there been different images of ducks the results would have been different. Machine learning is only a complex form of pattern recognition. The accuracy of what you get out is related to a) The quality of the learning data and b) The quality of the tested data when to try identification.

If your application of machine learning is safety critical and needs 100% accuracy, then machine learning might not be right for you.

Read about AI Machine Learning with Beacons