جامعة النهرين

Iraq’s water resources are increasingly affected by climate change, reduced transboundary river inflows, irregular hydrological events, and limited real-time monitoring infrastructure. These challenges make water allocation difficult, especially when decisions depend on fragmented data and delayed field measurements. This project proposes an artificial intelligence-assisted water resources management framework based on smart sensor networks, 26 GHz radar level sensors, and centralized data archiving. The system is designed around three main monitoring stations distributed along the Tigris and Euphrates Rivers and their connected lakes, with a central processing and archiving unit located in Rais Al-Bisha/Qurna. Real-time data on water level, quality, and regional demand are collected, transmitted, processed, and used to support adaptive water distribution decisions. A random sub-station verification mechanism is included to compare AI-generated decisions with live field readings before implementation. The expected outcome is a more reliable, responsive, and scalable decision-support system for improving water distribution, reducing losses, and supporting sustainable water management in Iraq.

This project presents an intelligent closed-loop prosthetic rehabilitation platform for lower-limb recovery. The system connects the prosthetic limb, artificial intelligence, EEG monitoring, VR training, and a mobile application into one adaptive rehabilitation workflow. The prosthetic sensing layer collects pressure, orientation, and ankle-motion data, while the AI module analyzes gait behavior and classifies the gait state as normal, imbalance, or abnormal. In parallel, EEG brainwave indicators are used to estimate attention and fatigue, allowing the system to respond not only to movement errors but also to the user’s readiness during training. The VR environment includes dedicated scenes for rehabilitation exercises, real-time sensor-reading visualization, corrective feedback, and relaxation. When EEG indicators show fatigue or reduced attention, the system automatically changes the active VR scene to relaxation mode, replacing the need for the patient to manually request a pause. The mobile application supports therapist monitoring by showing the patient’s condition, session progress, exercise history, pressure-distribution heatmaps, rehabilitation feedback, and patient–therapist chat. A pilot evaluation collected 11,533 multimodal records, achieving 92% recall, 94% precision, about 5% false-positive rate, and 453 ms mean end-to-end latency. The platform demonstrates a practical step toward safer, more personalized, and more engaging prosthetic rehabilitation.

يُعدّ فرز النفايات من الركائز الأساسية لتعزيز عمليات إعادة التدوير وتحقيق الاستدامة البيئية، إلا أن الاعتماد على الفرز اليدوي يؤدي إلى انخفاض الكفاءة وزيادة الوقت والجهد، فضلاً عن تعريض العاملين للمخاطر الصحية. يقدم هذا البحث نظامًا ذكيًا مدمجًا لفرز النفايات يعتمد على الحاسوب المصغر Raspberry Pi 5 وتقنيات التعلم العميق لإجراء عمليات التصنيف والفرز بصورة آلية وفي الزمن الحقيقي. يتكون النظام المقترح من كاميرا USB، ومستشعر للأشعة تحت الحمراء، ومستشعر للكشف عن المعادن، ومستشعر فوق صوتي، ومحرك مؤازر (Servo Motor)، ومحرك خطوي (Stepper Motor)، وجميعها مدمجة ضمن منصة مدمجة واحدة. وتم تدريب نموذج EfficientNetB4 باستخدام أسلوب Transfer Learning، ثم تحويله إلى TensorFlow Lite لتشغيله بكفاءة على منصة Raspberry Pi 5. يعتمد النظام على دمج نتائج التصنيف مع بيانات المستشعرات لاتخاذ قرار الفرز المناسب، حيث يتم توجيه النفايات تلقائيًا إلى الحاوية المخصصة بواسطة آلية ميكانيكية للتحكم بالحركة. أظهرت النتائج التجريبية أن النظام حقق دقة تصنيف بلغت حوالي 95%، مع زمن استدلال يتراوح بين 120–250 مللي ثانية، مما يؤكد قدرته على تنفيذ عمليات الفرز في الزمن الحقيقي بكفاءة عالية. ويُعد النظام المقترح حلاً منخفض الكلفة وقابلاً للتوسع، ويمكن الاستفادة منه في تطبيقات إدارة النفايات الذكية ودعم مبادرات المدن المستدامة.

This project presents the design, implementation, and evaluation of a heavy-duty multifunctional lawn mower system with remote control capabilities. The system combines a custom-fabricated octagonal steel chassis, four independently driven high-torque DC gear motors with planetary gearboxes, pneumatic off-road tires, and a 25.6 V / 100 Ah LiFePO4 battery pack, yielding reliable locomotion across grass, gravel, uneven soil, and moderate inclines. Control is implemented through a six-channel 2.4 GHz RF transmitter–receiver pair, allowing the operator independent management of drive direction, drive speed, blade activation, and emergency shut-off. Field testing demonstrated a top speed of 1.2 m/s on flat grass, an RF communication range of approximately 120 m in open field, a continuous runtime of 4.6 h under typical mowing load, and a mean cutting efficiency of 92 % across the test lawn. The proposed system offers a cost-effective and mechanically robust solution for residential and commercial lawn maintenance, particularly suited to the rough terrain and hot climate of the Middle East.

Dust accumulation on photovoltaic panels causes measurable efficiency losses that are particularly severe in arid environments. This paper presents an autonomous solar panel cleaning system that monitors panel performance directly through voltage and current sensors rather than relying on ambient dust detection. An ESP32 microcontroller computes instantaneous electrical power using P = V × I and transmits the result wirelessly to a smartphone via Bluetooth. A mobile application developed using MIT App Inventor applies an AI classification algorithm that evaluates incoming power data against a user-defined clean reference baseline, autonomously classifying the panel condition as Normal, Warning, or Danger, and triggering the cleaning cycle when the power drop reaches or exceeds 20%. Mechanically, a NEMA 23 stepper motor drives a rubber wiper across the panel surface while a water pump pre-wets the glass to improve dust removal. Outdoor testing in Baghdad, Iraq, demonstrated a voltage improvement from 16.95 V to 18.31 V after one cleaning cycle a gain of 8.02%. The complete prototype was assembled at a total cost of 179,500 IQD, confirming practical affordability for residential and small-scale solar maintenance applications.

This project presents the design and implementation of a dual-node distributed wireless smart glove system for real-time sign language translation. The proposed system aims to reduce communication barriers between deaf and hard-of-hearing individuals and the hearing community by converting dynamic sign language gestures into spoken words using embedded artificial intelligence. The system consists of two wearable smart gloves operating as transmitter and receiver nodes. Each glove integrates an ESP32 microcontroller, flex sensors, and an MPU6050 inertial measurement unit (IMU) to capture synchronized finger movements and hand orientation. Sensor data are exchanged through the low-latency ESP-NOW wireless communication protocol and combined into a unified data frame for processing. A pre-trained Long Short-Term Memory (LSTM) model deployed with TensorFlow Lite performs real-time gesture recognition, while a Text-to-Speech (TTS) module converts the recognized gestures into audible speech. The proposed architecture incorporates moving average filtering, frame validation, and confidence-based prediction mechanisms to improve recognition reliability and reduce false detections. Experimental evaluation demonstrates a response time of approximately 150 ms and a recognition accuracy exceeding 85%. As a lightweight, wearable, and cost-effective assistive device, the proposed system has potential applications in education, healthcare, customer service, and public accessibility, contributing to improved communication, social inclusion, and independence for deaf and hard-of-hearing individuals.

This project presents an intelligent dual-robot system that integrates a conveyor belt, computer vision, and Edge AI to automate object handling and sorting. Objects are transported by the conveyor belt and detected by an IR sensor, which stops the belt while an ESP32-S3-CAM identifies the object's shape. The first robot picks the object from the conveyor and transfers it to the second robot, which places it in the designated location according to its identified shape using synchronized control signals. The system achieves fast, accurate, and collision-free object sorting while improving automation efficiency and reducing processing latency.

This project addresses a practical problem in online retail for the Arabic-speaking market: conventional e-commerce search relies on exact keyword matching and therefore fails when a shopper types in colloquial Iraqi dialect, mixes Arabic and English, or expresses intent (a brand, a device type, or a price preference) rather than an exact product name. The proposed solution is a full-stack, three-tier platform that pairs a bilingual (Arabic/English) storefront with a dedicated hybrid search engine. The search engine fuses dense semantic retrieval (multilingual sentence-transformer embeddings stored in a vector database) with sparse lexical retrieval (BM25) using Reciprocal Rank Fusion, then applies an intent-aware re-ranking layer that understands brand, product-type, and price cues expressed in Arabic. The system is engineered for resilience: if the AI engine is unavailable, the storefront transparently falls back to the backend's SQL search so the site never loses its search capability. The outcome is a responsive, right-to-left–first shopping experience delivering relevant results for natural-language queries with low latency and graceful degradation.

NOVA هي منصة ذكاء اصطناعي متقدمة صُممت لتكون مساعداً إدارياً وتشغيلياً ذكياً للمؤسسات، من خلال دمج عدة وكلاء AI متخصصين داخل نظام واحد يعمل على تنظيم العمل، تحليل البيانات، أتمتة المهام، ودعم اتخاذ القرار بصورة أسرع وأكثر دقة. تعالج NOVA أحد أهم تحديات المؤسسات الحديثة، وهو تشتت المعلومات وضعف التنسيق بين الأقسام، حيث توفر المنصة وكلاء متخصصين في الإدارة، المحاسبة، التسويق، خدمة العملاء، والاجتماعات. يعمل كل وكيل على فهم البيانات والملفات المرتبطة بمجاله، ثم يقدم تحليلات، توصيات، تقارير، وتنبيهات تساعد الإدارة على تقليل الأخطاء، رفع الإنتاجية، وتحسين جودة القرارات. لا يقتصر دور NOVA على تنفيذ المهام فقط، بل يهدف إلى بناء بيئة عمل ذكية تحفظ معرفة المؤسسة، تربط بين الأقسام، وتحوّل البيانات اليومية إلى رؤى عملية قابلة للتنفيذ. وبذلك تمثل المنصة خطوة حقيقية نحو التحول الرقمي للمؤسسات الصغيرة والمتوسطة والجهات التعليمية والحكومية، من خلال حل عربي مرن، سهل الاستخدام، وقابل للتوسع مستقبلاً. NOVA ليست مجرد أداة مساعدة، بل نظام ذكي يساند المؤسسة في الإدارة، التخطيط، المتابعة، وتحسين الأداء، ليجعل العمل أكثر تنظيماً واحترافية واعتماداً على البيانات.

Ray is an AI-powered quadruped robotic cat designed to provide natural interaction using the Iraqi dialect and English. It combines an ESP32-S3 controller, a local AI server, voice recognition, and expressive movements to deliver an engaging and educational user experience. The robot supports both offline voice commands and AI-based conversations, making it an affordable and practical solution for home, educational, and research applications.

Res-Q AI is an AI-powered autonomous framework that transforms existing CCTV infrastructure into an intelligent emergency response network. The system detects traffic accidents in real time, eliminating delays caused by traditional eyewitness reporting and the associated social and legal barriers. Once an accident is detected, it automatically dispatches an alert with the exact location and guides emergency vehicles through the fastest route to the accident scene and the nearest hospital. By utilizing existing infrastructure without additional hardware, Res-Q AI provides a scalable, cost-effective solution for smart cities, enhancing emergency response efficiency, improving road safety, and helping save lives through faster medical intervention.

Sanabel is an integrated smart AgriTech platform for monitoring corn crop health in Iraqi agricultural fields, combining UAV technology, artificial intelligence, and IoT-based sensing into a unified system. The platform uses a Parrot ANAFI 4K drone to capture high-resolution aerial imagery of the field, which is analyzed through a YOLOv8-based AI model for early detection of corn diseases such as Northern Leaf Blight, Maize Streak Virus, and Gray Leaf Spot. The system integrates solar-powered ground stations equipped with RS485 soil sensors that measure humidity, temperature, and nutrient levels (NPK), communicating wirelessly via LoRa to a central Raspberry Pi 5 processing unit. Aerial imagery is further processed using WebODM to generate orthophoto maps and NDVI-based vegetation health analysis. Tested on a real corn field at the University of Baghdad, Sanabel achieved an 89.4% photogrammetric reconstruction success rate and identified Northern Leaf Blight as the dominant disease with 88.0% average detection confidence across the monitored field. The project aims to empower farmers with early disease detection, optimize water usage, and increase crop productivity, offering a practical and sustainable solution suited to Iraq's challenging environmental conditions.