GPU-Accelerated PD-IPM for Real-Time Model Predictive Control in Integrated Missile Guidance and Control Systems.

Area of Science:

• Aerospace Engineering
• Control Systems
• Computational Science

Background:

• Real-time model predictive control (MPC) is crucial for integrated guidance and control (IGC) in missile systems.
• The primal-dual interior point method (PD-IPM), a convex optimization technique, faces real-time performance challenges due to high computational costs, particularly in verifying Karush-Kuhn-Tucker (KKT) conditions.
• Existing methods struggle to meet the stringent real-time demands of advanced missile control systems.

Purpose of the Study:

• To develop and evaluate a graphics processing unit (GPU)-based method for accelerating PD-IPM in real-time MPC applications.
• To address the computational bottlenecks associated with KKT condition checks in PD-IPM for IGC systems.
• To enhance the real-time feasibility of MPC for missile guidance and control.

Main Methods:

• A novel approach utilizing GPU parallelization to accelerate the PD-IPM algorithm.
• Implementation of the GPU-accelerated PD-IPM on a standard embedded system for performance testing.
• Comparative analysis against conventional PD-IPM and other relevant control methods.

Main Results:

• The proposed GPU-based method significantly reduces the computation time required for PD-IPM.
• Demonstrated substantial improvements in real-time performance for MPC in IGC simulations.
• The method proved effective on a widely-used embedded system, validating its practical applicability.

Conclusions:

• GPU acceleration offers a viable solution to overcome the real-time limitations of PD-IPM in missile IGC.
• The proposed method enhances the computational efficiency and real-time control capabilities of missile systems.
• This advancement paves the way for more sophisticated and responsive guidance and control strategies.