Coursework

Coursework and related projects that I have completed during my Master's degree

In order to avoid hampering my professors who may re-use some of the assignments, I have only provided some examples of my programming work. Please reach out if you want to see more in detail!

CSCI 5302 - Advanced Robotics

Topics Learned: Robot Operating System (ROS), Kinematics and Controls, Motion Planning, Convex Optimization, Function Approximation, State Estimation (Bayes Filters, Kalman Filters, EKF, UKF), SLAM (Simultaneous Localization and Mapping), Partial Observability, Deep Reinforcement Learning, Safe Task and Motion Planning, Explainable AI, Intelligent Tutoring, Human-Agent Teaming, Collaborative AI
The final AWS Deepracer navigating a hallway, as fast as it can without crashing!

Assignments & Project

  • Implemented Rapidly-Exploring Random Trees (RRT) algorithm for motion planning in high-dimensional spaces, and enhanced it with RRT* to optimize paths and reduce travel distance, handling both holonomic and non-holonomic motion.
  • Configured ROS2 environment, explored nodes, topics, services, and actions. Developed and tested publisher/subscriber nodes and service/client nodes using Python.
  • Implemented value iteration and policy iteration algorithms for tabular cases in grid-world environments. Applied nearest-neighbor and n-linear interpolation methods for value iteration in continuous state spaces, specifically the MountainCar domain.
  • Utilized an AWS DeepRacer equipped with ROS2 Foxy on Ubuntu 20.04 to autonomously navigate a square hallway. Processed LIDAR data for obstacle detection and navigation, communicating with a PID controller via ROS2. Implemented camera data processing using YOLOv3 for stop sign recognition and response. Addressed challenges such as SLAM implementation, system memory management, and network communication issues.

CSCI 5622 - Machine Learning

Topics Learned: KNN, Linear Perceptrons, Linear & Logistic Regression, Deep Neural Networks, Explainable AI (LIME), Decision Trees & Random Forests, Dimensionality Reduction (PCA), Clustering, Ensemble Methods (Bagging, Boosting, Combination methods), Model Evaluation & Hyperparameter Tuning

Assignments

Sample
  • Analyzed breast cancer data to predict patient survival using K-Nearest Neighbors (K-NN) classifiers, adjusting distance metrics and K values to optimize prediction accuracy based on comprehensive performance metrics (Acc, BAcc, F1).
  • Developed a model to predict salinity levels from oceanographic measurements using linear regression, exploring feature correlations and impacts on model accuracy, and employed logistic regression for binary classification of salinity levels.
  • Implemented machine learning models for object recognition using the CIFAR-10 dataset, comparing the performance of Feedforward Neural Networks (FNNs) and Convolutional Neural Networks (CNNs) on image classification tasks. Also used LIME in order to explain the model.
  • Predicted job hireability from physiological and vocal measures during interviews using Decision Trees and Random Forests, analyzing model decision boundaries and ethical implications of automated hireability assessments.
  • Implemented speech-based ML model to detect depression and classify using XGBoost. Addressed class imbalances by using SMOTE, and mitigated gender bias by removing gender-dependent features.

ASEN/CSCI 5264 - Decision Making under Uncertainty

Topics Learned: Bayesian Networks, Markov Decision Processes (Value Iteration, Policy Iteration, Policy Evaluation), Reinforcement Learning (Model-based, Tabular, Policy Gradient, SARSA, Q-Learning), Neural Networks, Deep Q-Networks (DQN), Advanced Policy Gradient, Actor-Critic Methods, Entropy Regularization, Partially Observable Markov Decision Processes (Offline & Online solution Methods), Particle Filters, Basic Game Theory, Partially Observable Markov Games, Bayesian Network Learning, Imitation Learning, Inverse RL

Assignments

Sample
  • Implemented Online MDP Methods using Dense Grid World simulations; developed Monte Carlo Policy Evaluation and heuristic policies to enhance baseline performance. Conducted experiments with Monte Carlo Tree Search to optimize decision-making processes.
  • Engaged in the implementation and comparative analysis of tabular reinforcement learning algorithms including Q-Learning and SARSA, crafting learning curves to evaluate efficiencies and computational demands.
  • Modeled a POMDP for cancer monitoring using QuickPOMDPs.jl, executed policy evaluation with Monte Carlo methods, and integrated neural networks to approximate complex functions in reinforcement learning.
  • Developed QMDP and SARSOP solvers for the TigerPOMDP, contrasting heuristic and optimal policies within a cancer monitoring framework to evaluate efficacy and limitations of approximation methods.
  • Formulated high-performance policies and belief updaters for a Lasertag POMDP; employed a combination of heuristic approaches, deep reinforcement learning, and MCTS (Partially Observable Monte Carlo Policy) for adaptive decision-making capabilities.

ECEN 5612 - Random Processes

Topics Learned: Probability, Discrete Random Variables, Continuous Random Variables, Cumulative Distribution Functions, Bivariate Random Variables, Random Vectors, Minimum Mean-Square Error Estimation, Gaussian Random Vectors, Markov Processes, Poisson Processes, Wiener Processes, Linear Systems with Random Inputs, Wiener Filtering, Convergence of Random Sequences, Markov Chains

EMEN 5405 - Fundamentals of Systems Engineering

Topics Learned: The Systems Engineering Process, System Design Requirements, Engineering Design Methods and Tools, Design Review and Evaluation, System Engineering Program Planning, Organization for Systems Engineering, System Engineering Program Evaluation, Life Cycle of Systems, Requirements Analysis, Functional Analysis, Trade Studies, Risk Management, Integrated Product and Process Development (IPPD), Benchmarking Best Practices, Program Reporting and Feedback

Project

The class culminated with a final project where we worked in groups to design a system. We chose to design a targeting system for the Phalanx Block 1B Close-In Weapon System (CIWS) used by the U.S. Navy, focusing on integrating a sensor suite and display/control system to enhance target detection and tracking capabilities in various environmental conditions.

  • Created functional and physical block diagrams, allocated requirements to subsystems, and developed a specification compliance matrix to ensure all requirements were met.
  • Conducted a technology readiness assessment (TRA) for the tracking software, performed a trade study for the best display option, and identified key performance measures (TPM) to monitor design maturity.
  • Developed a detailed test plan, analyzed potential risks, and created a comprehensive program schedule and work breakdown structure (WBS) to manage the project timeline and tasks.