Project Topics & Ideas | Ashish Labade

Student Resources & Research

Project Topics & Ideas

A curated database of multidisciplinary engineering topic proposals and research directions across CSE, IoT, Mechatronics, and Cyber Security.

#	Topic Name	Abstract	Domain
1	Early Wildfire Ember Signature Detection using Multi-Spectral IR Pyrometers	This project deploys IoT nodes equipped with low-cost infrared pyrometers tuned to detect the specific thermal signature of burning embers (precursor to full ignition) rather than broad flames. A Python edge algorithm applies an Isolation Forest model to distinguish ember heat spikes from solar or engine thermal noise. The system fuses data from multiple nodes to track ember drift trajectories via wind data, providing firefighting teams with a 10-minute proactive warning window before a fire front expands.	IoT AI & ML Disaster Management
2	Dynamic EV Charging Load Forecasting via Induction Loop Vehicular Counters	This project repurposes existing in-road inductive loop sensors (used for traffic lights) to count passing vehicles and classify them (EV vs. ICE) using signature waveform analysis. A Python-based Spatio-Temporal Graph Network (ST-GNN) predicts the expected influx of EVs to nearby charging stations within the next hour. The system communicates with charging station controllers to dynamically adjust pricing and load shedding strategies, preventing grid overload.	IoT AI & ML Smart Mobility
3	Adversarial Water Distribution Leak Localization using Pressure Wave Velocimetry	This project places synchronized high-resolution pressure sensors at key junctions of a water pipeline network. When a leak occurs, it generates a negative pressure wave that propagates outward. The Python central server uses an optimized time-difference-of-arrival (TDOA) algorithm combined with a Physics-Informed Neural Network (PINN) to calculate the exact leak coordinates. Unlike basic flow monitoring, this system localizes leaks within meters, minimizing excavation and water loss.	IoT AI & ML Civil Infrastructure
4	Cold-Chain Perishable Shelf-Life Predictor via Volatile Organic Compound Sensing	This project integrates an array of MOX (Metal-Oxide) gas sensors to detect Volatile Organic Compounds (VOCs) emitted by ripening or spoiling produce inside cold-storage containers. Combined with humidity and temperature data, a Python-hosted Gradient Boosting model predicts the remaining shelf-life (in hours) of specific fruits/vegetables in real-time. The IoT gateway transmits alerts via LoRaWAN, allowing logistics managers to reroute perishable shipments proactively.	IoT AI & ML Logistics
5	Non-Intrusive Appliance Load Disaggregation (NILM) with Harmonic Fingerprinting	This project attaches a high-frequency current and voltage sensor to the main utility feed of a household. Using a Python-implemented Deep Residual Network, it disaggregates the aggregate power signal into individual appliance signatures by analyzing harmonic current distortions (up to the 50th harmonic). This unique approach identifies aging appliances (via degrading harmonics) and provides detailed, per-appliance energy breakdowns without needing smart plugs.	IoT AI & ML Smart Energy
6	Dynamic Urban Noise Cartography with Sound-Event Classification Nodes	This project deploys solar-powered IoT nodes with MEMS microphones across a city to classify urban sound events (e.g., construction, traffic, sirens, street music) using a lightweight 1D-CNN. Instead of just measuring decibels, the Python aggregation server fuses classified sound events with GPS timestamps to generate 4D (3D space + time) noise pollution cartography. The output helps urban planners identify transient noise hotspots and enforce dynamic zoning regulations.	IoT AI & ML Smart Cities
7	Acoustic Anomaly Interceptor for Rotary Machinery using Distributed Contact Mics	Instead of standard vibration sensors, this project uses piezoelectric contact microphones attached to multiple points on industrial motors to capture high-fidelity acoustic signatures. An Autoencoder neural network is trained on normal operational frequencies to detect subtle anomalous sounds indicating bearing wear or misalignment. The Python backend streams spectral data and triggers predictive maintenance alerts, significantly reducing false positives compared to traditional threshold-based systems.	IoT Industrial IoT (IIoT) AI & ML
8	IMU-Based Gait Asymmetry Predictor for Early Fall Risk Stratification	This wearable IoT system uses a 6-axis IMU (accelerometer + gyroscope) strapped to the lower back to capture gait dynamics during daily activities. A Python-based Temporal Convolutional Network (TCN) processes streaming sensor data to compute the Gait Symmetry Index (GSI) and detect micro-instabilities. Unlike basic step counters, this research project predicts fall risk scores weeks in advance, enabling proactive physiotherapeutic intervention.	IoT AI & ML Healthcare
9	Bio-Acoustic Pollinator Density Mapper using Wireless Sensor Arrays	This project places low-power microphone arrays across farmlands to capture high-frequency insect wingbeat sounds. An on-device Edge Impulse model running on Raspberry Pi classifies pollinator species (bees, wasps, flies) and estimates density in real-time. The aggregated data is sent to a central Python server, which generates a dynamic "pollination heatmap" to guide organic pesticide spraying schedules and assess crop fertilization risk.	IoT AI & ML Agri-Tech
10	Predictive HVAC Zoning via Occupancy Cognitive Fingerprinting	This project deploys distributed CO₂, PIR, and ambient light sensors to create an "occupancy cognitive fingerprint" for each room. Instead of binary detection, it uses a Transformer-based time-series model to forecast room-specific occupancy patterns 15 minutes ahead. The system proactively adjusts HVAC dampers and setpoints per zone, achieving energy optimization while maintaining thermal comfort, all coordinated via a Python-based edge gateway using MQTT.	IoT AI & ML Smart Buildings
11	Proactive Urban Flood Mapping via Real-Time Social Media Geotag Sentiment Fusion	This project creates a real-time flood monitoring dashboard by combining social media sentiment analysis with spatial data. It processes geotagged social media posts, analyzes sentiment related to flooding, and visualizes this data on a map. By fusing this human-centric data with remote sensing imagery, the system offers a proactive, comprehensive view of urban flooding, enabling faster emergency response and impact assessment.	AI & IoT
12	Agro-Climatic Yield Anomaly Predictor with Satellite Image Super-Resolution	This project uses Python and Google Earth Engine to process satellite data for agricultural monitoring. It employs super-resolution techniques (like data mining sharpeners) to enhance low-resolution imagery. By fusing this enhanced data with weather APIs, the system detects anomalies (e.g., drought stress, flooding) and predicts crop yield anomalies. This provides early, fine-scale warnings for food security and agricultural planning.	AI&ML Geospatial Analysis
13	Dynamic Supply Chain Stress-Test Simulator using Graph Neural Networks	This project models a supply chain as a complex, interdependent network using Graph Neural Networks (GNNs) with architectures like GCN and GAT. It builds a digital twin of the supply chain to simulate the cascading effects of disruptions (e.g., supplier shocks, demand spikes). The system "stress-tests" the network, predicts supplier risks, and provides actionable insights to enhance resilience against real-world disruptions.	AI & ML IoT
14	Generative Adversarial Network for Privacy-Preserving Synthetic Medical Time-Series Data	This project tackles the privacy constraints of healthcare data by using a Generative Adversarial Network (GAN) to create high-fidelity synthetic medical time-series data. It implements models like TimeGAN for sequential data and ADS-GAN to minimize patient re-identification risk. The system generates privacy-preserving datasets that maintain the statistical properties of real patient records, enabling secure data sharing and algorithm development without compromising sensitive information.	AI & ML Cybersecurity
15	Multimodal Emotion Dynamics Forecaster for Remote Mental Health Triage	This project develops a multimodal system that integrates facial expressions (via OpenCV), voice sentiment, and text analysis to track emotional states in real-time. By employing a multimodal transformer architecture and time-series forecasting, it models the temporal dynamics of emotions to identify trends and detect anomalies. The goal is to provide early warning signals for potential mental health crises, offering a data-driven triage tool for remote healthcare settings.	AI&ML Healthcare
16	Federated Learning-Enabled IoT Edge Framework for Real-Time Industrial Anomaly Detection and Predictive Maintenance	This system deploys lightweight federated learning models on edge devices attached to industrial machinery (CNC machines, motors, compressors) to detect anomalies from vibration, temperature, and acoustic sensors without sending raw data to the cloud. A React-based web dashboard visualises equipment health scores, remaining useful life estimates, and maintenance schedules. The aggregated global model is continuously improved using FedAvg across factory nodes.	AI & IoT