A Nonlinear Model Predictive Observer for Smart Projectile Applications
Abstract
The work reported here presents a nonlinear model predictive observer specialized to smart projectiles. The observer is based on a two-step process consisting of an initial state predictor followed by a state estimate corrector. We generate the predictor state by simply integrating the projectile's fixed plane equations of motion forward in time while the corrector step updates the state estimate by the minimization of the difference between measurements and model-predicted measurements over a finite duration of time in the past. Since the proposed method permits a general nonlinear representation of the projectile, the corrector step requires the solution of a nonlinear minimization problem which is subsequently solved with the use of a damped Newton procedure. The nonlinear model predictive observer is exercised on a smart projectile that employs a typical array of sensors including global positioning system, a three-axis gyroscope, three-axis accelerometers, and a three-axis magnetometer. Results with this smart weapon observer are promising since the observer is capable of estimating the full state of the projectile, including orientation and translational velocity, when realistic noise and bias errors are included in the sensor measurements. Relative to other existing observers, the proposed observer is computationally intensive because of the need to solve a nonlinear minimization problem at each computational cycle of the observer.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2008
- Accession Number
- ADA478966
Entities
People
- Mark Costello
- Ryan Letniak
Organizations
- Georgia Tech