Mixed-Horizon Retrospective-Predictive Control for Guaranteed Multibody Threat Engagement

Abstract

This proposed research consists of three main thrusts, namely, 1)input estimation for sensor/actuator failure detection, disturbance estimation, and target estimation; 2) closed-loop identification for parameter estimation, model estimation, and stability analysis; and 3) mixed-horizon retrospective-predictive control (MHRP). MHRP represents a fusion of adaptive control, which is primarily data-driven and uses minimal modeling information, and model predictive control (MPC), which is highly model driven but is able to accountfor constraints such as a terminal target set. The main innovations are 1) an extension of retrospective cost adaptive control (RCAC) to include soft constraints on the performance variable and 2) an output-feedback extension of MPC based on input-output models, which avoids the need for state observers. These extensions enable a fusion of RCAC and MPC, known as mixed-horizon retrospective-predictive control, which uses both the trailing horizon of RCACand the forward horizon of MPC. This fusion will take advantage of a set-valued modeling framework that uses zonotopes to propagate the output uncertainty set into the future. The overall methodology is facilitated by input estimation and closed-loop identification. All of the learning and control techniques that will be developed under this project are fully discrete (that is, digital) in both time and space, thus avoiding the mathematical and computational complexity of methods based on differential equations and facilitating the development andimplementation of computationally efficient numerical algorithms for on-board, real-time implementation.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 10, 2018
Source ID
N000141812211

Entities

People

  • Dennis S. Bernstein

Organizations

  • Board of Regents of the University of Michigan
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Neural Network Machine Learning.
  • Robotics and Automation.

Technology Areas

  • Space
  • Space - Space Objects
  • Space - Spacecraft Maneuvers