Open AccessCMAC Trained Optimum Mid course Guidance for Tactical Flight Vehicle
Author(s) -
Gulshan Kumar,
A. K. Sarkar
Publication year - 2021
Publication title -
defence science journal/defence science journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.198
H-Index - 32
eISSN - 0976-464X
pISSN - 0011-748X
DOI - 10.14429/dsj.71.16295
Subject(s) - pursuer , autopilot , robustness (evolution) , engineering , artificial neural network , control theory (sociology) , guidance system , terminal guidance , missile guidance , acceleration , simulation , computer science , control engineering , artificial intelligence , missile , aerospace engineering , law , biochemistry , chemistry , physics , control (management) , classical mechanics , political science , gene
This paper discusses design and validation of neural network based mid-course guidance law of a surface to air flight vehicle. In present study, initially different optimal trajectories have been generated off-line of different pursuer-evader engagements by ensuring minimum flight time, maximum terminal velocity and favorable handing over conditions for seeker based terminal guidance. These optimal trajectories have been evolved by nonlinear programming based direct method of optimisation. The kinematic information of both pursuer and evader, generated based on these trajectories have been used to train cerebellar model articulate controller (CMAC) neural network. Later for a given engagement scenario an on-line near optimal mid-course guidance law has been evolved based on output of trained network. Training has been carried out by CMAC type supervisory neural network. The tested engagement condition is within input/output training space of neural network. Seeker based homing guidance has been used for terminal phase. Complete methodology has been validated along pitch plane of pursuer-evader engagement. During mid-course phase, the guidance demand has been tracked by attitude hold autopilot and during terminal phase, the guidance demanded lateral acceleration has been tracked by acceleration autopilot. System robustness has been studied in presence of plant parameter variations and sensor noise under Monte Carlo Platform.