Automatic Differentiation

Author: Tianjiang Shuo

Website: https://cislunarspace.cn

Definition

A numerical technique that computes exact derivatives of a computer program by decomposing it into elementary operations and applying the chain rule. Unlike symbolic differentiation or finite differences, AD produces derivatives to machine precision without truncation error or amplified roundoff. Used in optimal control to efficiently generate sensitivity equations.

Application Value

This term has significant practical applications in cislunar mission planning and execution.

  • Intersection over Union
  • Point Spread Function
  • Flexible Image Transport System
  • Normalized Innovation Squared, NIS

References

  • Serban et al., 2002, Acta Astronautica
Improve this page
Last Updated:
Contributors: ouyangjiahong, Ou Yang Jiahong