• Computer Science > Robotics [Submitted on 14 Jan 2026] Title:TurboADMM: A Structure-Exploiting Parallel Solver for Multi-Agent Trajectory Optimization View PDF HTML (experimental)Abstract:Multi-agent trajectory optimization with dense interaction networks require solving large coupled QPs at control rates, yet existing solvers fail to simultaneously exploit temporal structure, agent decomposition, and iteration similarity. • One usually treats multi-agent problems monolithically when using general-purpose QP solvers (OSQP, MOSEK), which encounter scalability difficulties with agent count. • Structure-exploiting solvers (HPIPM) leverage temporal structure through Riccati recursion but can be vulnerable to dense coupling constraints. • We introduce TurboADMM, a specialized single-machine QP solver that achieves empirically near linear complexity in agent count through systematic co-design of three complementary components: (1) ADMM decomposition creates per-agent subproblems solvable in parallel, preserving block-tridiagonal structure under dense coupling; (2) Riccati warmstart exploits temporal structure to provide high-quality primal-dual initialization for each agent’s QP; (3) parametric QP hotstart \footnote{In the paper, we refer warmstart as the technique that uses the Riccati equation results as auxiliary QP initialization for a single QP solve, while hotstart as reusing the QR factorization across QP solve iterations.}in qpOASES reuses similar KKT system factorizations across
Article Summaries:
- Summary
A new solver, TurboADMM, has been introduced for multi‑agent trajectory optimization in robotics. The method tackles large, densely coupled quadratic programs (QPs) that arise when many agents interact over time. TurboADMM combines three complementary strategies: (1) an ADMM decomposition that splits the problem into per‑agent subproblems solvable in parallel while preserving block‑tridiagonal structure; (2) a Riccati‑based warm‑start that supplies high‑quality primal‑dual initializations for each agent’s QP; and (3) a parametric QP hot‑start that reuses KKT factorizations across ADMM iterations. Empirical results show near‑linear scaling with agent count, outperforming general‑purpose solvers such as OSQP and MOSEK on large‑scale scenarios.
Sources: