Non-linear dynamical system approach to behavior modeling

Siome Goldenstein, Edward Large, Dimitris Metaxas

Research output: Contribution to journalArticlepeer-review

23 Scopus citations


We present a dynamic systems approach to modeling and generating low-level behaviors for autonomous agents. Such behaviors include real-time target tracking and obstacle avoidance in time-varying environments. The novelty of the method lies on the integration of distinct non-linear dynamic systems to model the agent's interaction with the environment. An angular velocity control dynamic system guides the agent's direction angle, while another dynamic system selects the environmental input that will be used in the control system. The agent interacts with the environment through its knowledge of the position of stationary and moving objects. In our system agents automatically avoid stationary and moving obstacles to reach the desired target(s). This approach allows us to prove the stability conditions that result in a principled methodology for the computation of the system's dynamic parameters. We present a variety of real-time simulations that illustrate the power of our approach.

Original languageEnglish (US)
Pages (from-to)349-364
Number of pages16
JournalVisual Computer
Issue number7
StatePublished - 1999
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Software
  • Computer Vision and Pattern Recognition
  • Computer Graphics and Computer-Aided Design


Dive into the research topics of 'Non-linear dynamical system approach to behavior modeling'. Together they form a unique fingerprint.

Cite this