Direct neural dynamic programming

Jennie Si, Lei Yang, Derong Liu

Research output: Chapter in Book/Report/Conference proceedingChapter

14 Scopus citations


This chapter introduces direct neural dynamic programming (direct NDP), which belongs to the class of heuristic dynamic programming algorithms discussed in Chapters 3, 4, and 19. However, direct NDP is a model-independent approach to action-dependent heuristic dynamic programming. It is, therefore, an on-line learning control paradigm. This chapter contains a comparison study using other well-known algorithms to help readers gain quantitative insight on several ADP algorithms. It also contains results of direct NDP controlling a triple-link inverted pendulum using many continuous state variables and a continuous control, and direct NDP in a wireless network call admission control application. Furthermore, in Chapter 21 direct NDP is demonstrated on an industrial scale Apache helicopter model for stabilization, tracking control, and reconfiguration after component failure. Preliminary results indicate that direct NDP has the potential to address large-scale problems.

Original languageEnglish (US)
Title of host publicationHandbook of Learning and Approximate Dynamic Programming
PublisherWiley-IEEE Press
Number of pages27
ISBN (Electronic)9780470544785
ISBN (Print)047166054X, 9780471660545
StatePublished - Jan 1 2004


  • Adaptation model
  • Approximation methods
  • Artificial neural networks
  • Dynamic programming
  • Learning
  • Learning systems
  • Markov processes

ASJC Scopus subject areas

  • Computer Science(all)


Dive into the research topics of 'Direct neural dynamic programming'. Together they form a unique fingerprint.

Cite this