I’m happy to share that a paper resulting from my recent R&D work at ClearMotion, “Learning Feedforward Planners from Frequency-Domain Data,” has been accepted to the American Control Conference 2026, to be held in New Orleans.

The paper introduces a compact neural-network structure for generating continuous motion plans directly from frequency-domain data. The approach is fully data-driven and explicitly accounts for plant dynamics. It does not require system identification and greatly reduces the burden of manual controller tuning. The effectiveness of this method has been validated on ClearMotion vehicles.

I plan to present the work in person this May and will share a preprint of the final version in the coming weeks. Looking forward to engaging with colleagues in the controls community.