Tactical missiles are designed to target a variety of aerial and ground objects, which can be stationary or moving. To effectively intercept these targets with minimal miss-distance, tactical missiles require a highly efficient control system. Obtaining a high level of accuracy in the aerodynamic characterization of the missile is notably challenging. In addition, the difficulty in modeling external disturbances complicates the autopilot design. As a result, there is a significant need for a robust control system that can handle parametric uncertainties and unmodeled external disturbances. This paper presents a robust roll autopilot based on the ${{\mathcal{L}}_1}$ adaptive technique. The proposed roll autopilot is designed to handle uncertainties, external disturbances, and unmodeled sensor delays. This approach treats both external disturbances and parametric uncertainties within the roll loop as composite disturbances. A state observer is used to estimate the composite disturbance, and the system states in a unified manner. The estimates obtained from the inner loop uncertainties are used to calculate the adaptive roll control input, which enhances the baseline control input designed for nominal operating conditions. Simulation results are provided to demonstrate the effectiveness of the adaptive roll autopilot in achieving the specified design objectives.

Control of unstable systems is a pretty challenging task, since even a slight mismatch in controller gains can cause the system to go out of bounds. The control task becomes even more difficult if an unstable system has a zero in the right half-plane (RHP) of the complex s-plane. In this paper, a reinforcement learning (RL) based controller tuning method for non-minimum phase unstable systems is proposed. An adaptive PID controller is designed using RL and a modified Smith predictor structure is used to accommodate any delay in the system. The Deep Deterministic Policy Gradient (DDPG) algorithm has been employed along with a long short term memory (LSTM) layer in the actor as well as the critic network to design an RL-based adaptive PID controller. The proposed framework with the modified Smith predictor and RL based PID controller provide improved control performance. The performance has been evaluated against a state-of-the-art PID controller based on Integral Absolute Error (IAE) and Integral Square Error (ISE). Various case studies highlight the applicability of the proposed framework to a wide range of unstable and non-minimum phase systems.

The interaction effects among individual loops and multiple time delays are the prime reasons for degrading the closed-loop performance of multivariable systems. For highly interacting systems, decouplers are used to counteract interaction effects. However, decouplers may not be realizable and even if they are realizable, for high dimensional systems, decoupler elements may be complex. To improve the decoupler realizability, a dead-time compensation (DTC) approach is proposed in this work. The DTC framework can minimise the impact of dead-time on system dynamics. The proposed decoupler is designed based on the dynamic relative gain array (DRGA) to compensate for interaction effects. The DRGA reveals more insights about the coupling effect between a pair of input-output loops. The DRGA elements are used as decoupler elements for the multivariable process. The minimization of both interaction effects and dead-time dynamics can be achieved with the proposed DTC-based decoupling control approach. The simple IM C PI tuning rules are utilized to implement controllers for each loop. The simulation study affirming the proposed method provides better desired closed-loop performance, error indices for nominal and plant model mismatches, and robust stability against plant uncertainty over the popular existing decoupler based approaches.

A centralized PI control design methodology with easy applicability to small as well as large scale MIMO processes is proposed in this manuscript. The gain and phase margin tuning rule is deployed here to design the diagonal (D) PI controllers for the simple effective transfer function models (approximated for effective open-loop transfer functions). While accounting for the input-output coupling, the D controllers ensure robustness in the design along with good set-point tracking performance. The off-diagonal (OD) controllers in the centralized PI control framework are designed using the D PI controllers and the process steady-state gain. The OD controller design is carried out with an objective of minimizing the interaction effects. The performance of the proposed method is verified by simulation studies on various benchmark multivariable systems. Further, the robustness of the proposed method to uncertainties in the plant model is evaluated by performing closed-loop simulations subject to a plant-model mismatch of ±30%.