AE4800 Machine Learning for Autonomous Operations

This course covers the theory, computation and practical implementation of machine learning concepts for the design and autonomous operations of aerospace and marine engineering systems.  Specific machine learning topics covered are: construction of meta-models using three- and six-degree-of-freedom Newtonian dynamical equations, design of experiments for end-to-end dynamical systems, generation of labeled training data, neural networks, and deep learning.  These topics will be covered alongside the development of specific meta-models using physics-based simulations of aerospace and marine engineering systems. Examples will be drawn from a sample of DoD problems such as range prediction for a ballistic missile and predicting the damage from a simulated attack on an airfield.  Relevant concepts from optimization theory will also be covered as part of the foundations of machine learning.  MATLAB-based assignments will form the core of a student's learning experience.  This course may also be used as part of the allowable electives for the Robotics Certificate program.

Cross Listed Courses

Cross-listed with ME4800

Prerequisite

ME3420 OR AE3830 or by consent of instructor

Lecture Hours

3

Lab Hours

2

Quarter Offered

  • Summer

Outcomes

  1. Understand and explain the major deficiencies and challenges in machine learning in a mathematically justifiable manner with examples from aerospace/mechanical/marine engineering.
  2. Understand the scientific underpinnings of meta-models and their use in optimization and autonomous operations.
  3. Achieve proficiency at the engineering level of the mathematics of universal approximation, their limitations and their impacts on practical design optimization and autonomous operations.
  4. Understand first-hand the emerging tools and techniques in the application of machine learning for autonomous operations
  5. Achieve proficiency in implementation of machine learning concepts for design optimization and autonomous operations
  6. Demonstrate a flyable end-to-end algorithm for a platform of their choice.