Eiman Kanjo

Chieh-Tung Cheng, Mustafa Aslanov, Eiman Kanjo

Edge AI nodes for search and rescue are increasingly expected to run computer vision locally, yet ultra-low-end hardware imposes hard constraints on memory, compute, and inter-device communication. This work addresses occlusion-robust object detection on devices with less than 1 MB SRAM by combining an MCUNet backbone, a YOLOv2 detection head, and Lite quantisation. Two collaborative inference strategies are evaluated: feature-level fusion, concatenating intermediate feature maps, and decision-level fusion via Weighted Boxes Fusion (WBF). WBF outperforms feature-level fusion under all tested occlusion conditions, yielding gains of up to +0.2736 mAP in asymmetric scenarios. Extending fusion to three views improves accuracy further (up to +0.3827 mAP) at modest communication overhead (~1.3 KB per exchange). Hardware experiments progress from a host-assisted USB-relay baseline to a Wi-Fi peer-to-peer deployment on two Coral Dev Board Micro units, where WBF executes on-device with negligible communication energy relative to inference. In a 301.9 s autonomous session of 108 frames, fused output is produced on 61 frames versus 47 for a single board - a coverage gain of +29.8%. A decentralised federated learning feasibility note is included but not treated as a primary result, as performance remains limited under non-iid data. The results support decision-level fusion as a viable option for improving occlusion robustness in small-scale edge object detection, including host-free multi-board operation on ultra-low-end hardware.

Utkarsh Grover, Ravi Ranjan, Mingyang Mao et al.

Deploying foundation models in embodied edge systems is fundamentally a systems problem, not just a problem of model compression. Real-time control must operate within strict size, weight, and power constraints, where memory traffic, compute latency, timing variability, and safety margins interact directly. The Deployment Gauntlet organizes these constraints into eight coupled barriers that determine whether embodied foundation models can run reliably in practice. Across representative edge workloads, autoregressive Vision-Language-Action policies are constrained primarily by memory bandwidth, whereas diffusion-based controllers are limited more by compute latency and sustained execution cost. Reliable deployment therefore depends on system-level co-design across memory, scheduling, communication, and model architecture, including decompositions that separate fast control from slower semantic reasoning.

Eiman Kanjo, Mustafa Aslanov

The expansion of AI toward the edge increasingly exposes the cost and fragility of cen- tralised intelligence. Data transmission, latency, energy consumption, and dependence on large data centres create bottlenecks that scale poorly across heterogeneous, mobile, and resource-constrained environments. In this paper, we introduce Node Learning, a decen- tralised learning paradigm in which intelligence resides at individual edge nodes and expands through selective peer interaction. Nodes learn continuously from local data, maintain their own model state, and exchange learned knowledge opportunistically when collaboration is beneficial. Learning propagates through overlap and diffusion rather than global synchro- nisation or central aggregation. It unifies autonomous and cooperative behaviour within a single abstraction and accommodates heterogeneity in data, hardware, objectives, and connectivity. This concept paper develops the conceptual foundations of this paradigm, contrasts it with existing decentralised approaches, and examines implications for communi- cation, hardware, trust, and governance. Node Learning does not discard existing paradigms, but places them within a broader decentralised perspective

Xiaolu Fu, Ziyuan Bao, Eiman Kanjo

Vehicular ad hoc networks (VANETs) are a crucial component of intelligent transportation systems; however, routing remains challenging due to dynamic topologies, incomplete observations, and the limited resources of edge devices. Existing reinforcement learning (RL) approaches often assume fixed graph structures and require retraining when network conditions change, making them unsuitable for deployment on constrained hardware. We present TrajAware, an RL-based framework designed for edge AI deployment in VANETs. TrajAware integrates three components: (i) action space pruning, which reduces redundant neighbour options while preserving two-hop reachability, alleviating the curse of dimensionality; (ii) graph cross-attention, which maps pruned neighbours to the global graph context, producing features that generalise across diverse network sizes; and (iii) trajectory-aware prediction, which uses historical routes and junction information to estimate real-time positions under partial observations. We evaluate TrajAware in the open-source SUMO simulator using real-world city maps with a leave-one-city-out setup. Results show that TrajAware achieves near-shortest paths and high delivery ratios while maintaining efficiency suitable for constrained edge devices, outperforming state-of-the-art baselines in both full and partial observation scenarios.

Ziyuan Bao, Eiman Kanjo, Soumya Banerjee et al.

With the growing computational capabilities of microcontroller units (MCUs), edge devices can now support machine learning models. However, deploying decentralised federated learning (DFL) on such devices presents key challenges, including intermittent connectivity, limited communication range, and dynamic network topologies. This paper proposes a novel framework, bilayer Gossip Decentralised Parallel Stochastic Gradient Descent (GD PSGD), designed to address these issues in resource-constrained environments. The framework incorporates a hierarchical communication structure using Distributed Kmeans (DKmeans) clustering for geographic grouping and a gossip protocol for efficient model aggregation across two layers: intra-cluster and inter-cluster. We evaluate the framework's performance against the Centralised Federated Learning (CFL) baseline using the MCUNet model on the CIFAR-10 dataset under IID and Non-IID conditions. Results demonstrate that the proposed method achieves comparable accuracy to CFL on IID datasets, requiring only 1.8 additional rounds for convergence. On Non-IID datasets, the accuracy loss remains under 8\% for moderate data imbalance. These findings highlight the framework's potential to support scalable and privacy-preserving learning on edge devices with minimal performance trade-offs.

Qianxue Zhang, Eiman Kanjo

The advancement of technology has revolutionised the agricultural industry, transitioning it from labour-intensive farming practices to automated, AI-powered management systems. In recent years, more intelligent livestock monitoring solutions have been proposed to enhance farming efficiency and productivity. This work presents a novel approach to animal activity recognition and movement tracking, leveraging tiny machine learning (TinyML) techniques, wireless communication framework, and microcontroller platforms to develop an efficient, cost-effective livestock sensing system. It collects and fuses accelerometer data and vision inputs to build a multi-modal network for three tasks: image classification, object detection, and behaviour recognition. The system is deployed and evaluated on commercial microcontrollers for real-time inference using embedded applications, demonstrating up to 270$\times$ model size reduction, less than 80ms response latency, and on-par performance comparable to existing methods. The incorporation of the TinyML technique allows for seamless data transmission between devices, benefiting use cases in remote locations with poor Internet connectivity. This work delivers a robust, scalable IoT-edge livestock monitoring solution adaptable to diverse farming needs, offering flexibility for future extensions.

Stavros Mitsis, Ermos Hadjikyriakos, Humaid Ibrahim et al.

Deploying emotion recognition systems in real-world environments where devices must be small, low-power, and private remains a significant challenge. This is especially relevant for applications such as tension monitoring, conflict de-escalation, and responsive wearables, where cloud-based solutions are impractical. Multimodal emotion recognition has advanced through deep learning, but most systems remain unsuitable for deployment on ultra-constrained edge devices. Prior work typically relies on powerful hardware, lacks real-time performance, or uses unimodal input. This paper addresses that gap by presenting a hardware-aware emotion recognition system that combines acoustic and linguistic features using a late-fusion architecture optimised for Edge TPU. The design integrates a quantised transformer-based acoustic model with frozen keyword embeddings from a DSResNet-SE network, enabling real-time inference within a 1.8MB memory budget and 21-23ms latency. The pipeline ensures spectrogram alignment between training and deployment using MicroFrontend and MLTK. Evaluation on re-recorded, segmented IEMOCAP samples captured through the Coral Dev Board Micro microphone shows a 6.3% macro F1 improvement over unimodal baselines. This work demonstrates that accurate, real-time multimodal emotion inference is achievable on microcontroller-class edge platforms through task-specific fusion and hardware-guided model design.

Kieran Woodward, Eiman Kanjo, Georgios Papandroulidakis et al.

In recent years, the development of smart edge computing systems to process information locally is on the rise. Many near-sensor machine learning (ML) approaches have been implemented to introduce accurate and energy efficient template matching operations in resource-constrained edge sensing systems, such as wearables. To introduce novel solutions that can be viable for extreme edge cases, hybrid solutions combining conventional and emerging technologies have started to be proposed. Deep Neural Networks (DNN) optimised for edge application alongside new approaches of computing (both device and architecture -wise) could be a strong candidate in implementing edge ML solutions that aim at competitive accuracy classification while using a fraction of the power of conventional ML solutions. In this work, we are proposing a hybrid software-hardware edge classifier aimed at the extreme edge near-sensor systems. The classifier consists of two parts: (i) an optimised digital tinyML network, working as a front-end feature extractor, and (ii) a back-end RRAM-CMOS analogue content addressable memory (ACAM), working as a final stage template matching system. The combined hybrid system exhibits a competitive trade-off in accuracy versus energy metric with $E_{front-end}$ = $96.23 nJ$ and $E_{back-end}$ = $1.45 nJ$ for each classification operation compared with 78.06$μ$J for the original teacher model, representing a 792-fold reduction, making it a viable solution for extreme edge applications.

Thomas Johnson, Eiman Kanjo, Kieran Woodward

In this paper, we define the term 'DigitalExposome' as a conceptual framework that takes us closer towards understanding the relationship between environment, personal characteristics, behaviour and wellbeing using multimodel mobile sensing technology. Specifically, we simultaneously collected (for the first time) multi-sensor data including urban environmental factors (e.g. air pollution including: PM1, PM2.5, PM10, Oxidised, Reduced, NH3 and Noise, People Count in the vicinity), body reaction (physiological reactions including: EDA, HR, HRV, Body Temperature, BVP and movement) and individuals' perceived responses (e.g. self-reported valence) in urban settings. Our users followed a pre-specified urban path and collected the data using a comprehensive sensing edge devices. The data is instantly fused, time-stamped and geo-tagged at the point of collection. A range of multivariate statistical analysis techniques have been applied including Principle Component Analysis, Regression and spatial visualisations to unravel the relationship between the variables. Results showed that EDA and Heart Rate Variability HRV are noticeably impacted by the level of Particulate Matters (PM) in the environment well with the environmental variables. Furthermore, we adopted Deep Belief Network to extract features from the multimodel data feed which outperformed Convolutional Neural Network and achieved up to (a=80.8%, σ=0.001) accuracy.

Kieran Woodward, Eiman Kanjo, Athanasios Tsanas

The quantification of emotional states is an important step to understanding wellbeing. Time series data from multiple modalities such as physiological and motion sensor data have proven to be integral for measuring and quantifying emotions. Monitoring emotional trajectories over long periods of time inherits some critical limitations in relation to the size of the training data. This shortcoming may hinder the development of reliable and accurate machine learning models. To address this problem, this paper proposes a framework to tackle the limitation in performing emotional state recognition on multiple multimodal datasets: 1) encoding multivariate time series data into coloured images; 2) leveraging pre-trained object recognition models to apply a Transfer Learning (TL) approach using the images from step 1; 3) utilising a 1D Convolutional Neural Network (CNN) to perform emotion classification from physiological data; 4) concatenating the pre-trained TL model with the 1D CNN. Furthermore, the possibility of performing TL to infer stress from physiological data is explored by initially training a 1D CNN using a large physical activity dataset and then applying the learned knowledge to the target dataset. We demonstrate that model performance when inferring real-world wellbeing rated on a 5-point Likert scale can be enhanced using our framework, resulting in up to 98.5% accuracy, outperforming a conventional CNN by 4.5%. Subject-independent models using the same approach resulted in an average of 72.3% accuracy (SD 0.038). The proposed CNN-TL-based methodology may overcome problems with small training datasets, thus improving on the performance of conventional deep learning methods.

Kieran Woodward, Eiman Kanjo, David J Brown et al.

Children can find it challenging to communicate their emotions especially when experiencing mental health challenges. Technological solutions may help children communicate digitally and receive support from one another as advances in networking and sensors enable the real-time transmission of physical interactions. In this work, we pursue the design of multiple tangible user interfaces designed for children containing multiple sensors and feedback actuators. Bluetooth is used to provide communication between Tangible Toys (TangToys) enabling peer to peer support groups to be developed and allowing feedback to be issued whenever other children are nearby. TangToys can provide a non-intrusive means for children to communicate their wellbeing through play.

Thomas Johnson, Eiman Kanjo, Kieran Woodward

The growth of mobile sensor technologies have made it possible for city councils to understand peoples' behaviour in urban spaces which could help to reduce stress around the city. We present a quantitative approach to convey a collective sense of urban places. The data was collected at a high level of granularity, navigating the space around a highly popular urban environment. We capture people's behaviour by leveraging continuous multi-model sensor data from environmental and physiological sensors. The data is also tagged with self-report, location coordinates as well as the duration in different environments. The approach leverages an exploratory data visualisation along with geometrical and spatial data analysis algorithms, allowing spatial and temporal comparisons of data clusters in relation to people's behaviour. Deriving and quantifying such meaning allows us to observe how mobile sensing unveils the emotional characteristics of places from such crowd-contributed content.

Kieran Woodward, Eiman Kanjo, David J. Brown et al.

Stress is a growing concern in modern society adversely impacting the wider population more than ever before. The accurate inference of stress may result in the possibility for personalised interventions. However, individual differences between people limits the generalisability of machine learning models to infer emotions as people's physiology when experiencing the same emotions widely varies. In addition, it is time consuming and extremely challenging to collect large datasets of individuals' emotions as it relies on users labelling sensor data in real-time for extended periods. We propose the development of a personalised, cross-domain 1D CNN by utilising transfer learning from an initial base model trained using data from 20 participants completing a controlled stressor experiment. By utilising physiological sensors (HR, HRV EDA) embedded within edge computing interfaces that additionally contain a labelling technique, it is possible to collect a small real-world personal dataset that can be used for on-device transfer learning to improve model personalisation and cross-domain performance.

Kieran Woodward, Eiman Kanjo, Andreas Oikonomou

In recent years, machine learning has developed rapidly, enabling the development of applications with high levels of recognition accuracy relating to the use of speech and images. However, other types of data to which these models can be applied have not yet been explored as thoroughly. Labelling is an indispensable stage of data pre-processing that can be particularly challenging, especially when applied to single or multi-model real-time sensor data collection approaches. Currently, real-time sensor data labelling is an unwieldy process, with a limited range of tools available and poor performance characteristics, which can lead to the performance of the machine learning models being compromised. In this paper, we introduce new techniques for labelling at the point of collection coupled with a pilot study and a systematic performance comparison of two popular types of deep neural networks running on five custom built devices and a comparative mobile app (68.5-89% accuracy within-device GRU model, 92.8% highest LSTM model accuracy). These devices are designed to enable real-time labelling with various buttons, slide potentiometer and force sensors. This exploratory work illustrates several key features that inform the design of data collection tools that can help researchers select and apply appropriate labelling techniques to their work. We also identify common bottlenecks in each architecture and provide field tested guidelines to assist in building adaptive, high-performance edge solutions.

Ben Nicholls, Chee Siang Ang, Eiman Kanjo et al.

Mindless eating, or the lack of awareness of the food we are consuming, has been linked to health problems attributed to unhealthy eating behaviour, including obesity. Traditional approaches used to moderate eating behaviour often rely on inaccurate self-logging, manual observations or bulky equipment. Overall, there is a need for an intelligent and lightweight system which can automatically monitor eating behaviour and provide feedback. In this paper, we investigate: i) the development of an automated system for detecting eating behaviour using wearable Electromyography (EMG) sensors, and ii) the application of such a system in combination with real time wristband haptic feedback to facilitate mindful eating. Data collected from 16 participants were used to develop an algorithm for detecting chewing and swallowing. We extracted 18 features from EMG and presented those features to different classifiers. We demonstrated that eating behaviour can be automatically assessed accurately using the EMG-extracted features and a Support Vector Machine (SVM): F1-Score=0.94 for chewing classification, and F1-Score=0.86 for swallowing classification. Based on this algorithm, we developed a system to enable participants to self-moderate their chewing behaviour using haptic feedback. An experiment study was carried out with 20 additional participants showing that participants exhibited a lower rate of chewing when haptic feedback delivered in forms of wristband vibration was used compared to a baseline and non-haptic condition (F (2,38)=58.243, p<0.001). These findings may have major implications for research in eating behaviour, providing key new insights into the impacts of automatic chewing detection and haptic feedback systems on moderating eating behaviour with the aim to improve health outcomes.

Kieran Woodward, Eiman Kanjo, David Brown

Mental well-being technologies possess many qualities that give them the potential to help people receive assessment and treatment who may otherwise not receive help due to fear of stigma or lack of resources. The combination of advances in sensors, microcontrollers and machine learning is leading to the emergence of dedicated tangible interfaces to monitor and promote positive mental well-being. However, there are key technical, ergonomic and aesthetic challenges to be overcome in order to make these interfaces effective and respond to users' needs. In this paper, the barriers to develop mental well-being tangible interfaces are discussed by identifying and examining the recent technological challenges machine learning, sensors, microcontrollers and batteries create.User-oriented challenges that face the development of mental well-being technologies are then considered ranging from user engagement during co-design and trials to ethical and privacy concerns.