Module 9
Machine Learning: Overview
Real Case Study
References
Machine Learning: Overview
AI / ML Fundamentals
Artificial Intelligence (AI)
Artificial Intelligence (AI) is a branch of computer science that focuses on creating systems capable of performing tasks that normally require human intelligence. These tasks include learning, reasoning, problem-solving, decision-making, language understanding, visual perception, and pattern recognition. AI is widely used in robotics, healthcare, finance, autonomous systems, aerospace, smart assistants, and industrial automation.
Machine Learning (ML)
Machine Learning (ML) is a subfield of Artificial Intelligence that enables computers to learn from data, identify patterns, and make predictions or decisions without being explicitly programmed for every task. Instead of following fixed rules, ML systems improve their performance as they are exposed to more data. Machine learning is widely applied in recommendation systems, speech recognition, image classification, fraud detection, predictive maintenance, and autonomous navigation.
AI Vs ML
| Comparison | Artificial Intelligence (AI) | Machine Learning (ML) |
|---|---|---|
| Definition | A broad field that enables machines to mimic human intelligence. | A subset of AI that allows machines to learn from data. |
| Scope | Includes reasoning, planning, NLP, robotics, computer vision, and learning. | Primarily focused on learning patterns and making predictions from data. |
| Goal | To create intelligent systems capable of decision-making and automation. | To develop models that improve performance using experience (data). |
| Approach | May use rules, logic, search methods, or learning algorithms. | Relies mainly on statistical models and algorithms trained on data. |
| Examples | Chatbots, autonomous robots, expert systems, smart assistants. | Spam filtering, image recognition, demand forecasting, recommendation systems. |
| Complexity | Broader and more comprehensive domain. | More specialized and data-driven branch within AI. |
Types of Machine Learning
The three major types of Machine Learning are:
1. Supervised Learning – Learning from labeled data where the correct output is already known.
2. Unsupervised Learning – Learning from unlabeled data to discover hidden patterns, structures, or groupings.
3. Reinforcement Learning – Learning by interacting with an environment and receiving rewards or penalties based on actions.
1. Supervised Learning – Learning from labeled data where the correct output is already known.
2. Unsupervised Learning – Learning from unlabeled data to discover hidden patterns, structures, or groupings.
3. Reinforcement Learning – Learning by interacting with an environment and receiving rewards or penalties based on actions.
Supervised Learning
Supervised Learning is a type of machine learning in which the model is trained using labeled datasets. Each training example contains an input and its corresponding correct output. The algorithm learns the relationship between input variables and target outputs so that it can predict outcomes for new unseen data.
Supervised learning is mainly used for:
• Classification – Predicting categories (e.g., spam or not spam, defect or no defect)
• Regression – Predicting numerical values (e.g., temperature, cost, demand)
It is one of the most widely used forms of machine learning in real-world applications such as medical diagnosis, aircraft fault detection, weather prediction, and financial forecasting.
Supervised learning is mainly used for:
• Classification – Predicting categories (e.g., spam or not spam, defect or no defect)
• Regression – Predicting numerical values (e.g., temperature, cost, demand)
It is one of the most widely used forms of machine learning in real-world applications such as medical diagnosis, aircraft fault detection, weather prediction, and financial forecasting.
Unsupervised Learning
Unsupervised Learning is a type of machine learning in which the model works with unlabeled data. Since no correct outputs are provided, the algorithm attempts to find hidden structures, patterns, clusters, or relationships within the dataset.
Unsupervised learning is mainly used for:
• Clustering – Grouping similar data points together
• Dimensionality Reduction – Reducing the number of features while preserving useful information
• Pattern Discovery – Identifying trends and hidden structures in data
It is useful in customer segmentation, anomaly detection, image compression, and exploratory data analysis.
Unsupervised learning is mainly used for:
• Clustering – Grouping similar data points together
• Dimensionality Reduction – Reducing the number of features while preserving useful information
• Pattern Discovery – Identifying trends and hidden structures in data
It is useful in customer segmentation, anomaly detection, image compression, and exploratory data analysis.
Reinforcement Learning
Reinforcement Learning is a type of machine learning where an agent learns by interacting with an environment. The agent performs actions, receives feedback in the form of rewards or penalties, and gradually improves its strategy to maximize long-term rewards.
Reinforcement learning is commonly used in:
• Robotics and autonomous vehicles
• Game-playing AI systems
• Navigation and control systems
• Industrial process optimization
It is highly valuable for decision-making problems where actions influence future outcomes.
Reinforcement learning is commonly used in:
• Robotics and autonomous vehicles
• Game-playing AI systems
• Navigation and control systems
• Industrial process optimization
It is highly valuable for decision-making problems where actions influence future outcomes.
Supervised Vs Unsupervised Machine Learning
| Comparison | Supervised Learning | Unsupervised Learning |
|---|---|---|
| Data Type | Uses labeled data | Uses unlabeled data |
| Objective | Learn input-output mapping | Discover hidden patterns or structures |
| Main Tasks | Classification and Regression | Clustering and Dimensionality Reduction |
| Output Availability | Known during training | Unknown during training |
| Evaluation | Easy to evaluate using metrics like accuracy, precision, recall, RMSE | Harder to evaluate because no true labels exist |
| Examples | Email spam detection, house price prediction, disease diagnosis | Customer segmentation, anomaly detection, pattern grouping |
Algorithms of Supervised ML
Commonly used supervised machine learning algorithms include:
• Linear Regression
• Multiple Linear Regression
• Logistic Regression
• Support Vector Machine (SVM)
• Decision Tree
• Random Forest
• k-Nearest Neighbors (k-NN)
• Naïve Bayes Classifier
• Gradient Boosting Machine (GBM)
• XGBoost
• Artificial Neural Networks (ANN)
• Convolutional Neural Networks (CNN)
• Linear Regression
• Multiple Linear Regression
• Logistic Regression
• Support Vector Machine (SVM)
• Decision Tree
• Random Forest
• k-Nearest Neighbors (k-NN)
• Naïve Bayes Classifier
• Gradient Boosting Machine (GBM)
• XGBoost
• Artificial Neural Networks (ANN)
• Convolutional Neural Networks (CNN)
Algorithms of Unsupervised ML
Commonly used unsupervised machine learning algorithms include:
• K-Means Clustering
• Hierarchical Clustering
• DBSCAN (Density-Based Spatial Clustering of Applications with Noise)
• Gaussian Mixture Models (GMM)
• Principal Component Analysis (PCA)
• Independent Component Analysis (ICA)
• t-Distributed Stochastic Neighbor Embedding (t-SNE)
• UMAP (Uniform Manifold Approximation and Projection)
• K-Means Clustering
• Hierarchical Clustering
• DBSCAN (Density-Based Spatial Clustering of Applications with Noise)
• Gaussian Mixture Models (GMM)
• Principal Component Analysis (PCA)
• Independent Component Analysis (ICA)
• t-Distributed Stochastic Neighbor Embedding (t-SNE)
• UMAP (Uniform Manifold Approximation and Projection)
ML Frameworks & Libraries
Several frameworks and libraries are used for machine learning model development, training, deployment, and inference:
TensorFlow – A widely used open-source framework for developing and training machine learning and deep learning models.
TensorFlow Lite – A lightweight version of TensorFlow optimized for mobile, embedded, and edge devices.
PyTorch – A powerful deep learning framework commonly used in research, experimentation, and model development due to its flexibility and dynamic computation graph.
OpenCV – An open-source computer vision library used for image processing, feature extraction, object detection, and video analysis.
ONNX Runtime – A high-performance inference engine that enables machine learning models to run efficiently across multiple platforms and hardware systems.
Scikit-learn – A popular Python library used for classical machine learning algorithms such as classification, regression, clustering, and dimensionality reduction.
TensorFlow – A widely used open-source framework for developing and training machine learning and deep learning models.
TensorFlow Lite – A lightweight version of TensorFlow optimized for mobile, embedded, and edge devices.
PyTorch – A powerful deep learning framework commonly used in research, experimentation, and model development due to its flexibility and dynamic computation graph.
OpenCV – An open-source computer vision library used for image processing, feature extraction, object detection, and video analysis.
ONNX Runtime – A high-performance inference engine that enables machine learning models to run efficiently across multiple platforms and hardware systems.
Scikit-learn – A popular Python library used for classical machine learning algorithms such as classification, regression, clustering, and dimensionality reduction.
Applications of AI / ML
AI and ML are widely applied across different industries and engineering domains. Common applications include:
• Image recognition and computer vision
• Natural language processing and chatbots
• Recommendation systems
• Predictive maintenance
• Fraud detection and cybersecurity
• Medical diagnosis and healthcare analytics
• Autonomous vehicles and drones
• Aerospace system monitoring and fault prediction
• Weather forecasting and climate analytics
• Smart manufacturing and Industry 4.0
• Image recognition and computer vision
• Natural language processing and chatbots
• Recommendation systems
• Predictive maintenance
• Fraud detection and cybersecurity
• Medical diagnosis and healthcare analytics
• Autonomous vehicles and drones
• Aerospace system monitoring and fault prediction
• Weather forecasting and climate analytics
• Smart manufacturing and Industry 4.0
Real Case Study
Key Applications of ML in UAV Engineering
Machine Learning has become a transformative technology in UAV engineering by enabling drones to operate more intelligently, safely, and autonomously. It improves mission planning, decision-making, perception, navigation, and system reliability in both civilian and defense applications.
Commonly used applications include:
Object Detection and Tracking – Identifying and continuously following vehicles, humans, animals, or infrastructure targets from aerial imagery.
Obstacle Avoidance – Detecting trees, buildings, power lines, and moving obstacles to prevent collisions during autonomous flight.
Navigation and Path Prediction – Predicting safe and efficient flight paths based on terrain, weather, and mission constraints.
Fault Detection and Health Monitoring – Monitoring sensors, motors, batteries, ESCs, and flight systems to detect abnormalities before failure occurs.
Precision Agriculture – Analyzing crop health, irrigation patterns, weed detection, and field monitoring using multispectral or RGB imagery.
Terrain Mapping – Processing aerial images and sensor data to create maps, digital elevation models, and 3D terrain reconstructions.
Anomaly Detection – Identifying unusual patterns in flight behavior, system logs, surveillance footage, or environmental data.
Swarm Drone Coordination – Enabling multiple UAVs to coordinate, communicate, distribute tasks, and operate collaboratively in real time.
Commonly used applications include:
Object Detection and Tracking – Identifying and continuously following vehicles, humans, animals, or infrastructure targets from aerial imagery.
Obstacle Avoidance – Detecting trees, buildings, power lines, and moving obstacles to prevent collisions during autonomous flight.
Navigation and Path Prediction – Predicting safe and efficient flight paths based on terrain, weather, and mission constraints.
Fault Detection and Health Monitoring – Monitoring sensors, motors, batteries, ESCs, and flight systems to detect abnormalities before failure occurs.
Precision Agriculture – Analyzing crop health, irrigation patterns, weed detection, and field monitoring using multispectral or RGB imagery.
Terrain Mapping – Processing aerial images and sensor data to create maps, digital elevation models, and 3D terrain reconstructions.
Anomaly Detection – Identifying unusual patterns in flight behavior, system logs, surveillance footage, or environmental data.
Swarm Drone Coordination – Enabling multiple UAVs to coordinate, communicate, distribute tasks, and operate collaboratively in real time.
Software Required to Integrate ML Approaches
Software integration for machine learning-based UAV systems involves a combination of operating systems, middleware, flight control software, development frameworks, optimization tools, and simulation platforms. These tools help in building, testing, deploying, and validating intelligent UAV solutions.
Operating System
• Ubuntu / Linux-based environments provide a stable, developer-friendly, and ML-compatible platform for robotics and AI applications.
Middleware and Communication
• ROS / ROS2 (Robot Operating System) enables communication between sensors, controllers, perception modules, and machine learning nodes.
• MAVLink is a lightweight communication protocol used for interfacing mission logic, telemetry, and autonomous control with UAV hardware.
Flight Control Firmware
• PX4 is an open-source flight control stack used in autonomous drones and UAV research.
• ArduPilot supports fixed-wing, multirotor, rover, and VTOL systems with extensive autonomous capabilities.
Machine Learning Frameworks
• TensorFlow / TensorFlow Lite
• PyTorch
• OpenCV
• ONNX Runtime
• Scikit-learn
These libraries support data processing, model training, deployment, image analytics, and real-time inference.
Embedded and Edge AI Tools
• NVIDIA JetPack
• CUDA
• TensorRT
• Edge Impulse
• OpenVINO
These tools are used to optimize ML models for onboard deployment on embedded systems such as Jetson, Intel, or edge AI processors.
Simulation and Training Platforms
• Gazebo
• AirSim
• MATLAB / Simulink
• Unreal Engine
These platforms allow safe testing of control logic, perception systems, mission autonomy, and ML models before real-world flight implementation.
Operating System
• Ubuntu / Linux-based environments provide a stable, developer-friendly, and ML-compatible platform for robotics and AI applications.
Middleware and Communication
• ROS / ROS2 (Robot Operating System) enables communication between sensors, controllers, perception modules, and machine learning nodes.
• MAVLink is a lightweight communication protocol used for interfacing mission logic, telemetry, and autonomous control with UAV hardware.
Flight Control Firmware
• PX4 is an open-source flight control stack used in autonomous drones and UAV research.
• ArduPilot supports fixed-wing, multirotor, rover, and VTOL systems with extensive autonomous capabilities.
Machine Learning Frameworks
• TensorFlow / TensorFlow Lite
• PyTorch
• OpenCV
• ONNX Runtime
• Scikit-learn
These libraries support data processing, model training, deployment, image analytics, and real-time inference.
Embedded and Edge AI Tools
• NVIDIA JetPack
• CUDA
• TensorRT
• Edge Impulse
• OpenVINO
These tools are used to optimize ML models for onboard deployment on embedded systems such as Jetson, Intel, or edge AI processors.
Simulation and Training Platforms
• Gazebo
• AirSim
• MATLAB / Simulink
• Unreal Engine
These platforms allow safe testing of control logic, perception systems, mission autonomy, and ML models before real-world flight implementation.
AI/ML-based UAV Mission
An AI/ML-based UAV mission refers to a drone operation in which machine learning and artificial intelligence techniques are integrated into the mission workflow to enable intelligent sensing, autonomous decision-making, adaptive control, and data-driven mission execution.
In a conventional UAV mission, the drone typically follows pre-programmed routes and fixed logic. In contrast, an AI/ML-enabled mission allows the UAV to perceive the environment, interpret sensor inputs, identify mission targets, adapt to uncertainties, and optimize actions in real time.
A typical AI/ML-based UAV mission may include the following stages:
1. Mission Planning – Defining objectives such as surveillance, mapping, agriculture, inspection, or delivery, while preparing flight paths and operational constraints.
2. Sensor Data Acquisition – Collecting real-time data from RGB cameras, thermal cameras, LiDAR, GPS, IMU, ultrasonic sensors, or multispectral payloads.
3. Intelligent Perception – Using AI/ML models to detect objects, classify targets, identify terrain features, or recognize anomalies from onboard sensor data.
4. Autonomous Decision-Making – Making intelligent decisions such as route correction, obstacle avoidance, target following, or mission reprioritization based on live inputs.
5. Navigation and Control – Integrating ML outputs with flight controllers to guide the UAV safely and efficiently during operation.
6. Mission Execution – Performing the assigned task such as crop scanning, infrastructure inspection, terrain mapping, search and rescue, or surveillance.
7. Post-Processing and Analysis – Analyzing mission data for reporting, visualization, anomaly assessment, predictive insights, and future mission optimization.
AI/ML-based UAV missions are highly valuable because they improve autonomy, operational safety, target accuracy, adaptability, and mission intelligence while reducing human workload and reaction time.
In a conventional UAV mission, the drone typically follows pre-programmed routes and fixed logic. In contrast, an AI/ML-enabled mission allows the UAV to perceive the environment, interpret sensor inputs, identify mission targets, adapt to uncertainties, and optimize actions in real time.
A typical AI/ML-based UAV mission may include the following stages:
1. Mission Planning – Defining objectives such as surveillance, mapping, agriculture, inspection, or delivery, while preparing flight paths and operational constraints.
2. Sensor Data Acquisition – Collecting real-time data from RGB cameras, thermal cameras, LiDAR, GPS, IMU, ultrasonic sensors, or multispectral payloads.
3. Intelligent Perception – Using AI/ML models to detect objects, classify targets, identify terrain features, or recognize anomalies from onboard sensor data.
4. Autonomous Decision-Making – Making intelligent decisions such as route correction, obstacle avoidance, target following, or mission reprioritization based on live inputs.
5. Navigation and Control – Integrating ML outputs with flight controllers to guide the UAV safely and efficiently during operation.
6. Mission Execution – Performing the assigned task such as crop scanning, infrastructure inspection, terrain mapping, search and rescue, or surveillance.
7. Post-Processing and Analysis – Analyzing mission data for reporting, visualization, anomaly assessment, predictive insights, and future mission optimization.
AI/ML-based UAV missions are highly valuable because they improve autonomy, operational safety, target accuracy, adaptability, and mission intelligence while reducing human workload and reaction time.
AI/ML-based UAV Mission Dashboard
A fully responsive UAV dashboard for Aerospace Education, Drone Design Studies, UAV Mission Classification, STEM / R&D Dashboards, and Student Internship Projects. This platform categorizes UAVs based on AI/ML research domains and suggests mission-relevant configurations, sensors, hardware, software, algorithms, and design logic.
Showing all UAVs and AI/ML categories.
References
Reference
1. S.S. Chin. Missile Configuration Design. McGraw Hill, 1961.
2. Mark Pinney Aerodynamics of Missiles and Rockets. McGraw-Hill Education, 2013.
3. Marvin Hobbs Fundamentals of Rockets, Missiles, and Spacecraft. J.F. Rider, 1962.
4. Sethunathan, P., Sugendran, R. N., & Anbarasan, T. Aerodynamic Configuration design of a missile at Int J Eng Res & Technol (IJERT), 2015.
5. Jack N. Nielsen Missile Aerodynamics. NIELSEN ENGINEERING & RESEARCH, INC, 1988.
6. Siouris, George Missile Guidance and Control Systems. Springer New York, 2006.
2. Mark Pinney Aerodynamics of Missiles and Rockets. McGraw-Hill Education, 2013.
3. Marvin Hobbs Fundamentals of Rockets, Missiles, and Spacecraft. J.F. Rider, 1962.
4. Sethunathan, P., Sugendran, R. N., & Anbarasan, T. Aerodynamic Configuration design of a missile at Int J Eng Res & Technol (IJERT), 2015.
5. Jack N. Nielsen Missile Aerodynamics. NIELSEN ENGINEERING & RESEARCH, INC, 1988.
6. Siouris, George Missile Guidance and Control Systems. Springer New York, 2006.
UAV Machine Learning by Dr Aishwarya Dhara