An Optimal End-Game Guidance Law with a Jerk Constraint
By Natan Grinfeld and Joseph Z. Ben-Asher
Many tactical and strategic missiles have been using Proportional Navigation Guidance (PNG) law for decades. Its simple implementation along with its high accuracy has made PN the most popular guidance law in many applications. Some specific applications, however, may face problems when trying to implement the PNG law. One such application is the case of a physical constraint on the jerk (first derivative of the acceleration) for the intercept missile. This case may occur mainly in exo-atmospheric, thrust vector maneuvering missiles, in which the generation of the required acceleration maneuver is less immediate then the case of an endo-atmospheric aerodynamic maneuvering missiles. Though being brought into consideration indirectly through the time constant of the flight control, the jerk constraint has never been implemented directly and explicitly when deriving an optimal guidance law. The need for such a development of an optimal, jerk constraint, guidance law arises when introducing such constraint on the jerk on a common PNG law. It was first shown, as motivation for the study, that under a tight jerk constraint the missile misses the target using PN. This research objective is to develop an end-game phase optimal guidance law with jerk limit considerations and investigate its performance in comparison to PN and APN (Augmented Proportional Navigation) laws and in the cases of a physical constraint on the jerk. By introducing a fourth state variable of m n
(missile's acceleration) and adding a "price" on its derivative, thus obtaining the system of 1 2 3 4 ; ; ; t m x y x y x n x n = = = = , and solving the problem of: ( ) 2 1 0, min 2 f o t f m t y t J n dt | | = = | \ . } , using Cauchy-Schwartz inequality, an optimal, jerk minimizing guidance law was developed. The new guidance command is in the form of: 3 2 : 1 10; ; 2 m go go go go f N ZEM n t where N ZEM y t y t y t t t ' = ' = = + + =
The new guidance law, named JLG (Jerk Limited Guidance), was tested against both PN and APN. The results are interesting: JLG turns out to be more efficient than PN under a jerk limit as expected, and under certain jerk limitations, the missile hit the target using JLG while missing using PN. Since it is not a fair game comparing JLG to PN because JLG is using knowledge of the targets maneuver while PN isnt, it was decided to compare JLG to APN which also uses that knowledge. When comparing those methods against a constant target acceleration, both methods hit the target but with a very different behavior. While APN reacts to the targets maneuver with an instant acceleration peak in order to cancel the targets advantage (while doing so, assuming the target will continue with the same maneuver), JLG has a much more moderate acceleration reaction, building its acceleration slowly, to arrive at a maximum level of acceleration only after half of the end game time. This encouraged the test of both methods against a bang-bang maneuvering target. It turns out that against a bang-bang maneuvering target, under a jerk limit, JLG is much more efficient, at some cases (depending on the acceleration switch time) hitting the target while APN misses, thus due to the moderate reaction which is much more suited to changes in the maneuver than the straight forward approach of APN. Next to be studied in the research, is the guidance law performance and sensitivity to changes through an adjoint system representation.