The Flight Dynamics Control Lab (FDCL) at KAIST, led by Professor Chang-Hun Lee, has developed a customized solver for real-time onboard convex optimization, addressing the computational challenges of deploying optimization-based guidance and control (G&C) in embedded systems.

As autonomous systems and their mission requirements grow in complexity, there is an increasing need for control strategies that can effectively achieve mission objectives while satisfying operational constraints. Optimization-based G&C has emerged as a powerful framework to address this challenge, as demonstrated in applications such as SpaceX’s Falcon 9 landing burn. However, these approaches require significant computational resources, necessitating the development of optimization solvers that are both efficient and robust for reliable onboard operation.

To address this challenge, Prof. Lee’s research group developed a customized convex optimization solver tailored for real-time onboard applications. The solver is built on a predictor–corrector primal–dual interior-point framework combined with a homogeneous embedding approach, enabling detection of infeasibility while maintaining high computational efficiency.

Unlike conventional methods, the proposed solver can directly handle quadratic objective second-order cone programming (SOCP) problems without requiring reformulation, thereby preserving problem structure and avoiding unnecessary computational overhead. In addition, the research team developed a code generation tool that automatically analyzes problem sparsity and generates customized solver code. The generated solver is implemented in C with minimal dependencies and supports fully static memory allocation that facilitates flight software validation.

The performance of the developed solver was validated through benchmark tests against existing methods, demonstrating consistent outperformance across various problem sizes and conditions in onboard environments. This demonstrates the practical feasibility of the developed solver for deployment in real-world autonomous systems.

Further technical details can be found in the published paper: J.-I. Jang and C.-H. Lee, “Customized Interior-Point Methods Solver for Embedded Real-Time Convex Optimization,” Accepted for publication in IEEE Transactions on Aerospace and Electronic Systems, 2026.

This work was supported in part by the Theater Defense Research Center funded by Defense Acquisition Program Administration under Grant UD240002SD and in part by the Korea Aerospace Research Institute (KARI) through its own research project titled “Development of Control Performance Analysis Program for Spaceplanes”.
2026 KI Newsletter