Talk: IEEE CDC 2025 (Rio) — Intrinsic Successive Convexification

I presented our work at IEEE CDC 2025 in Rio de Janeiro, Brazil.

What problem was this about?

Successive convexification is a powerful trajectory optimization tool, but its standard formulation can become representation-dependent and computationally redundant on manifold-valued state spaces.

What was the main contribution?

We introduced intrinsic successive convexification (iSCvx), where linearization, perturbations, and state updates are done intrinsically on the manifold (via tangent spaces and retractions), not in redundant ambient coordinates.

Why it matters

iSCvx is representation-invariant, keeps iterates on the manifold, and reduces unnecessary dimensions in local subproblems—important for scalable optimization in aerospace and robotics guidance problems.

• 📊 Presentation
• 📄 Paper PDF
• 🔗 IEEE Xplore