Jan Kamm, Antonio Terpin and Raffaello D’Andrea on motion planning for a million robots

Jan Kamm, Antonio Terpin and Raffaello D’Andrea on motion planning for a million robots

Authors Jan Kamm, Antonio Terpin and Raffaello D’Andrea present sequential contact insertion (SCI), a method for planning collision-free motion by organizing optimization around moments when robots reach their minimum allowed separation. Rather than discretizing and optimizing entire trajectories, SCI starts with unconstrained trajectories, inserts contacts where separation is violated, and adjusts contact times and states; trajectories between contacts are reconstructed analytically. The authors report numerical experiments planning for one million robots in minutes and thousands in fractions of a second. On problems feasible for a fixed-grid sequential convex programming baseline, they report median speedups of several orders of magnitude at matching solution accuracy. The results address a scaling problem in multi-agent motion planning: collision-avoidance constraints couple robot trajectories and make the optimization nonconvex, while optimizing whole trajectories becomes expensive for large fleets. The abstract describes numerical experiments, not deployment with physical robot fleets; how the method performs in such settings remains an open question.

Publication: 17ebeaea-04be-492b-84f9-797055df1421

attribution — pass: The reading correctly credits Jan Kamm, Antonio Terpin and Raffaello D'Andrea as the authors. It keeps the method and results attributed to them using phrasing such as 'present' and 'report'.

Authors:Jan Kamm, Antonio Terpin, Raffaello D'Andrea

summary — uncertain: The reading is faithful, but central terms go unexplained for a general reader. It never says what 'nonconvex' implies (a much harder optimization) or what a 'fixed-grid sequential convex programming baseline' is, yet the speedup claim depends on that comparison. It also omits the abstract's real-world context, such as drone shows and autonomous operations. It drops a scope condition: planning uses simplified double-integrator dynamics, with low-level controllers handling physical execution. Smallest repair: add tooltips or plain glosses for these terms and note the planning-level modeling assumption.

robot motion is often modeled using double-integrator dynamics (with low-level controllers handling the physical dynamics during execution)

date — pass: originalPublished is 2026-10-30. This matches the arXiv v1 submission date and comes before the feed date of 2026-11-02.

[Submitted on 30 Oct 2026]

support — pass: All four claims match the abstract verbatim, and their scope conditions are kept: the speedups apply only to baseline-feasible problems and are median figures. The open question about physical fleets is framed as unknown rather than asserted. No benefits or costs are invented.

On problems feasible for a fixed-grid sequential convex programming baseline, SCI achieves median speedups of several orders of magnitude at matching solution accuracy.

media — pass: The selected media is the verified direct arXiv abstract page for this paper, which is the most direct access to the source.