Research

Papers & publications.

  • Counter-UAV vision Critical review 2026

    Optical Flow Methods for Interceptor Drones: A Critical Review

    Saeed Althabahi, Alanoud Alteneiji

    A critical review of 47 works (2021–2026) on optical-flow motion perception for autonomous counter-UAV interception — spanning classical, CNN (FlowNet, PWC-Net), RAFT, transformer (FlowFormer), edge, and event-camera methods, benchmarked by accuracy, speed, size, and power for SWaP-constrained drones.

  • Edge AI · Drones Systematic review 2026

    Edge AI and TinyML for Vision Applications in Autonomous Drones: A Systematic Critical Review

    Saeed Althabahi, Sultan Alafeefi

    A systematic review of Edge AI and TinyML for on-board drone vision across five domains — comparing model compression (quantization, pruning, distillation) and accelerators (Jetson, Coral, Hailo-8, Loihi 2) under the weight, power, and thermal limits of UAV payloads.

  • Autonomous agents Survey 2025

    Bridging Strategic Reasoning and Tactical Execution with LLM and RL Agents

    Saeed Althabahi

    A survey synthesizing hierarchical architectures that pair LLM strategic planners with RL tactical controllers. Hybrid systems report ~28% higher task success and better sample efficiency than RL-only baselines on long-horizon autonomous-control tasks.

  • Smart environments Review 2026

    Self-Managing Edge AI Systems for Intelligent Environments

    Saeed Althabahi

    A review of self-managing Edge AI for intelligent environments — smart homes, buildings, and tactical edge networks — covering model compression, runtime optimization, reinforcement learning, and federated learning for autonomous, resource-aware operation.

  • Predictive maintenance Applied study 2025

    Leveraging Lightweight AI for Anomaly Detection in Mechanical Power Transmission at the Edge

    Walid Ayadi, Amar Amouri, Saeed Althabahi, Nada Alzahmi, M. N. Khalid

    A TinyML model on an Arduino Nano 33 BLE Sense with an IMU detects DC-motor faults (loose, misaligned, noisy) from vibration signals — reaching 97.1% validation accuracy with millisecond on-device inference for low-cost predictive maintenance.

  • Medical AI Journal paper 2025

    AI-Powered CNN Model for Automated Lung Cancer Diagnosis in Medical Imaging

    Walid Ayadi, Yasser Farhat, Saeed Althabahi, et al.
    Int. J. of Statistics in Medical Research (2025)

    A convolutional neural network for automated lung-cancer classification on the IQ-OTHNCCD CT dataset (benign / malignant / normal), achieving 95% accuracy and a 0.95 F1-score. Peer-reviewed and published open access.