Abstract: It is known that if a nonlinear control affine system without drift is bracket generating, then its associated sub-Laplacian is invertible under some conditions on the domain. In this note, we investigate the converse. We show how invertibility of the sub-Laplacian operator implies a weaker form of controllability, where the reachable sets of a neighborhood of a point have full measure. From a computational point of view, one can then use the spectral gap of the (infinite-dimensional) self-adjoint operator to define a notion of degree of controllability. An essential tool to establish the converse result is to use the relation between invertibility of the sub-Laplacian to the controllability of the corresponding continuity equation using possibly non-smooth controls. Then using Ambrosio–Gigli–Savaré’s superposition principle from optimal transport theory we relate it to controllability properties of the control system. While the proof can be considered of the Perron–Frobenius type, we also provide a second dual Koopman point of view.
Introduction
In nonlinear control theory, determining whether a system can be steered between arbitrary states is a fundamental challenge. Traditionally, this is addressed using the Lie Algebra Rank Condition (LARC), which evaluates whether the system's control directions can span the entire state space through Lie bracket combinations. While powerful, LARC is primarily a local test—it guarantees controllability within a small neighborhood but offers limited insight into global behavior. Furthermore, LARC provides a binary result: a system is either controllable or it is not. It lacks a metric to quantify the "degree" of controllability or the effort required for steering.
This research addresses these limitations by investigating the converse of a known relationship: if a system is bracket-generating, its associated sub-Laplacian operator is invertible. The paper asks whether the invertibility of this linear operator can serve as a global test for controllability and if its spectral properties can quantify control performance.
Methodology: From Trajectories to Ensembles
The core innovation lies in shifting the perspective from individual trajectories (governed by ODEs) to the evolution of probability densities or "ensembles" (governed by PDEs). This approach employs several key mathematical frameworks:
- The Sub-Laplacian ($\Delta_H$): A linear operator is defined to measure function variations along allowed control directions. Invertibility of this operator suggests that the available controls have sufficient "reach" to influence the entire domain.
- Dual Perspectives: The study explores the Perron–Frobenius view, which tracks the evolution of probability densities, and the Koopman view, which focuses on the evolution of observables (functions of the state).
- Superposition Principle: To connect smooth PDE solutions with potentially irregular nonlinear trajectories, the author utilizes the Ambrosio–Gigli–Savaré superposition principle from optimal transport theory. This allows the results to hold even when control laws are non-smooth.
Key Insights: Linearizing the Operator, Not the System
Standard linearization techniques often discard the nonlinear dynamics (like Lie bracket effects) that enable controllability. By instead "linearizing" at the operator level—using the sub-Laplacian—the nonlinear structure is preserved within a linear algebraic framework.
A significant insight is the introduction of the spectral gap ($\lambda$) as a measure of controllability. Much like the Controllability Gramian in linear systems, the spectral gap of the sub-Laplacian provides a quantitative "degree of controllability." A larger gap indicates a more easily controlled system, while a vanishing gap suggests the emergence of unreachable regions.
Principal Results
The theoretical investigation yields two primary findings:
- Full Measure Reachability (Theorem 3): If the sub-Laplacian $\Delta_H$ is invertible on a given domain, then for any target neighborhood, the set of starting points capable of reaching that neighborhood has full Lebesgue measure. This implies the system is "almost everywhere" controllable.
- Quantifiable Controllability (Corollary 4): The author demonstrates that if a system satisfies a Poincaré inequality with a constant $\lambda > 0$, it is approximately controllable. This $\lambda$ serves as a computable metric for the degree of controllability.
These results suggest that controllability can be tested using convex optimization and numerical PDE solvers, potentially simplifying the analysis of complex nonlinear systems such as advanced robotics or chemical processes.
Limitations and Future Directions
While providing a robust global test, the current framework has specific constraints:
- Driftless Systems: The results are currently limited to control-affine systems without drift (systems that remain stationary without input). Systems with inherent momentum or drift present significantly higher complexity.
- Approximate vs. Exact: The findings guarantee approximate controllability (reaching a neighborhood) and full-measure reachability, but do not yet guarantee exact point-to-point controllability in all cases.
- Domain Geometry: The proofs rely on specific technical assumptions regarding the domain boundary (e.g., $\epsilon-\delta$ subdomains). Extending these results to more general or pathological domains remains an open area for research.
Summary updated