Application of Evolutionary Computing in Control Allocation

This issue of interaction of control allocation and actuator dynamics and has been dealt with by very few researchers. What was not considered in most control allocation algorithms is the fact that the control surfaces are manipulated by either hydraulics or electric actuators, and constitute a dynamic system which cannot produce infinite accelerations. In other words, if a control was initially at rest, and later commanded to move at its maximum rate in some direction for a specified amount of time, it would gradually build up speed until it reached the commanded rate. The final position of the control would therefore not be the same as that calculated using the commanded rate and the time during which it was instructed to move (Bolling 1997). In this chapter, a method, which post-processes the output of a control allocation algorithm, is developed to compensate for actuator dynamics. The method developed is solved for a diagonal matrix of gain corresponding to individual actuators. This matrix is then multiplied with the commanded change in control effector settings as computed by the control allocator and actuators dynamics interactions. The basic premise of this method is to post process the output of the control allocation algorithm to overdrive the actuators so that at the end of a sampling interval the actual actuator positions are equivalent to the desired actuator positions (Oppenheimer and Doman 2004). The overdriving of the actuators is done by multiplying the change in commanded signal with the identified gain matrix which is called the compensator. This identification is done by using a soft computing technique (i.e. genetic algorithms). The simulation setup including control allocator block, compensator and actuator rig makes a non-linear set up. During the identification of the compensator using this setup by soft computing technique such as genetic algorithms, the likelihood of the solution being a global minimum is high as compared to other optimisation techniques. This is why genetic algorithms have been used in this analysis rather than other techniques such as linear programming. The main contribution is to design a compensator using an evolutionary computing technique (i.e. genetic algorithms) to compensate the interaction between control allocation and actuator dynamics. It should be mentioned that in this method the model of the actuator does not need to be known. The simulation setup consists of excitation signals, the control allocation block, the compensator and the actuators rig. When designing control allocation typically the actuator dynamics are ignored because the bandwidth of the actuators is larger than the frequencies of the rigid body modes of the aircraft. Fig. 1 shows a control allocator with actuator dynamics neglected. If there is a case