New control framework uses sliced optimal transport for distribution steering

Researchers have developed a new control framework for distribution steering, a method that guides the state law of a dynamical system between specified initial and terminal distributions. This framework leverages sliced optimal transport, which simplifies the process by using one-dimensional projections instead of full state-space constructions. The developed method results in a randomized controller that, when averaged, provides a deterministic sliced feedback, effectively steering the system towards the target distribution while minimizing control energy. This approach has been demonstrated to work for single-integrator dynamics and linear dynamical systems, preserving Gaussianity and steering mean and covariance values for Gaussian endpoint laws. AI

IMPACT This research introduces a novel control method for dynamical systems, potentially impacting areas that require precise state distribution manipulation.

RANK_REASON This is a research paper detailing a new mathematical framework and control method. [lever_c_demoted from research: ic=1 ai=0.4]

Read on arXiv stat.ML →

• Distribution Steering
• Gaussian function
• optimal transport
• single-integrator dynamics
• Sliced Optimal Transport
• Sliced Optimal Transport Control
• Wasserstein metric

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →