Hungu Lee

Generalized input-estimation technique for tracking maneuvering targets

A new input estimation technique for target tracking problem is proposed. Conventional input estimation techniques assume that the target maneuver level is constant within the detection window, which has been the major drawback of the techniques. The proposed technique is developed to overcome this drawback by modeling the target maneuver as a linear combination of some basic time functions. The resulting algorithm has a generalized formulation including earlier works on input estimation. A detection performance of the proposed algorithm is analyzed by investigating the detection sensitivity according to the selection of maneuver models and other design parameters such as the detection window size, measurement noise level, and sampling step size. A computer simulation study shows that the estimation performance of the proposed algorithm is comparable to Boglers input estimation method while the computation time is greatly reduced.

Acceleration of the convergence speed of evolutionary algorithms using multi-layer neural networks

Despite the global optimization capability and low sensitivity to initial parameter estimates, evolutionary algorithms suffer from heavy computational loads especially when the fitness evaluation is time-consuming. The proposed acceleration method implements an online multi-layer neural network approximating the fitness calculation, which greatly decreases the computation time because the time-consuming fitness calculation can be replaced by the simple network output. The acceleration is achieved as the number of individuals used for the network training gradually decreases according to an adaptive scheme. A convergence theorem guarantees convergence to the optimal solution as well as ensuring the network stability. The proposed method is verified by a numerical example.

Control design of spinning rockets based on co-evolutionary optimization

Abstract This paper deals with an attitude control-system design technique for a spinning sounding rocket. A proportional, integral, and derivative (PID) type controller is developed for a dynamic model simplified by a complex summation method. The optimal PID gains at each design point are automatically obtained by using the co-evolutionary augmented Lagrangian method which can handle complicated specifications and constraints. The performance of the control system is verified through six degrees-of-freedom non-linear simulations, which shows that the optimized control system is robust against plant uncertainties and disturbances.

On-line suboptimal midcourse guidance using neural networks

This paper considers mideourse guidance of Air to-air tactical missiles where seeker lock-on is not Achieved at launch. An on-line suboptimal midcourse guidance law, which is a neural network approximation of optimal feedback strategy, is derived to eliminate the need for solving the two-point boundary value problems in real time. Since the proposed guidance law is an approximation, it is not very accurate but suitable for midcourse guidance. The approach is to teach the neural network to extract the information about the mapping in question from the off line generated optimal trajectories corresponding to various final conditions in the region where a target is predicted to appear and to use it as a feedback scheme for real-time on-board implementation. It needs an assumption that the optimal control signal can be expressed as a function of the states and final conditions.

Parameter robust control design using bimatrix co-evolution algorithms

The parameter robust controller design problem is often expressed as a game between the controller and uncertain parameters. The game is expressed as a constrained minimax problem if design requirements are treated as the constraints, but the saddle-point solution of the constrained minimax problem is difficult to obtain by using conventional methods. This paper proposes a new method of designing parameter robust controllers by introducing a bimatrix co-evolution algorithm which solves the constrained minimax problem. The proposed algorithm approximates the problem as a bimatrix game between two groups having independent fitness measures and finds the saddle-point solution through successive bimatrix games. The resulting controller guarantees robustness against bounded parameter uncertainties and satisfies the design requirements. The performance of the proposed algorithm is verified by its application to a satellite attitude control problem.

Missile guidance using neural networks

Abstract This paper introduces a new guidance algorithm using neural networks for bank-to-turn (BTT) missiles. The proposed guidance algorithm compensates for the missile dynamics by using the inverse dynamics learned by neural networks. The new guidance law is applied to a full-order nonlinear BTT missile model, and the performance is compared with that of the proportional navigation guidance law.

Modified Mendel operation for multimodal function optimization

Applying Mendels hereditary law to genetic algorithms opened new possibilities. A recent application of Mendels law to a global optimization problem showed promising results. However, its definition of Mendel operations were not quite clear. We clarify and redefine the Mendel operation. We also propose two methods to modify the Mendel operation for better performance in solving a multimodal function problem. The modified Mendel operation is used on a simple multimodal function to see the effects.

Fault Tolerant Attitude Control for a Spacecraft Using Reaction Wheels

This paper considers a fault tolerant control problem for a spacecraft using reaction wheels. Faults are assumed to be inherent to only actuators(reaction wheels) and a control algorithm to accommodate actuators` faults is proposed. An attitude control loop includes an angular velocity control loop. The time delay control method is used to make a spacecraft follow the command angular velocity and to accommodate actuators` faults. A stability condition for the proposed algorithm is derived and the performance is demonstrated by computer simulations.

Papers : Feasibility Study on Attitude Control of Spacecraft Using Pulsed Plasma Thrusters

In this paper, the feasibility of the attitude control of a spacecraft using pulsed plasma thrusters(PPTs) is studied. The PPT consumes less propellant mass requied for the orbit management or attitude control owing to its high specific impulse characteristics, compared with traditional gas propulsion system. The PPT is expected to be highly adequete for the missions requiring long-duration operations because it has relatively long operation time and easy implementation. The feasibility of the PPT for attitude control of a small satellite system is addressed through realistic missions. The classical PD controller and a fuzzy logic controller are tested, and fuel saving fuzzy logic controller is then proposed for more flexible mission performance.

Suboptimal Guidance for Passive Homing Missile With Angle-Only-Measurement

Abstract The optimal guidance law obtained from a nonlinear stochastic problem may have a dual-control effect. The guidance law with dual-control effect is addressed by a trade-off study by maintaining good guidance performance and small estimation errors. The direct stochastic optimization is solved by the co-evolutionary augmented Lagrangian method. Because the solution is an open-loop type for various engagement scenarios, a neural-network is implemented to derive a feedback guidance law. Simulation results show that the proposed guidance law is superior to the conventional proportional navigation.