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-Savare'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
Determining the controllability of nonlinear systems—the ability to steer a state from any point A to any point B—remains one of the most significant challenges in control theory. Traditionally, researchers rely on the Lie Algebra Rank Condition (LARC), which evaluates the "brackets" of vector fields to determine local reachability. However, LARC is fundamentally a local condition and often proves computationally intractable when applied globally to complex, high-dimensional systems.
In linear systems theory, the Controllability Gramian provides an elegant, global certificate of reachability. This paper proposes a nonlinear analogue by shifting the perspective from individual trajectories (the ODE view) to the evolution of probability densities (the PDE view). By analyzing the spectral properties of a specific linear operator—the sub-Laplacian—we can certify global controllability properties through a convex, linear test.
The Sub-Laplacian Framework
The core of this approach is the sub-Laplacian operator, $\Delta_H$, defined based on the system's control vector fields. This operator is linear and self-adjoint, allowing us to leverage functional analysis to study nonlinear dynamics.
The Primal View: Perron-Frobenius and Optimal Transport
The "primal" proof strategy views the control system as a mechanism for pushing forward ensembles of particles. If the sub-Laplacian is invertible, one can construct control laws that move one probability density to another via the continuity equation.
To bridge the gap between moving abstract densities and moving physical points, the paper employs the Ambrosio-Gigli-Savaré superposition principle from optimal transport theory. This principle ensures that if a "cloud" of points can be moved from one region to another in the density space, there must exist valid underlying ODE trajectories that perform this work, even if the required control signals are non-smooth.
The Dual View: Koopman Operator
As a dual perspective, the author utilizes the Koopman operator framework. This approach demonstrates that if a system is not controllable—meaning it possesses invariant sets that the state cannot escape—the sub-Laplacian cannot be injective. Thus, the presence of a non-trivial kernel in the operator serves as a direct indicator of a lack of global controllability.
Key Insights: Controllability as a Spectrum
1. The Converse Result
While it was previously established that "bracket generating" systems (those satisfying LARC) possess invertible sub-Laplacians, this work proves the more critical converse: Invertibility of the sub-Laplacian implies approximate global controllability. This provides a sufficient condition that can be tested using linear methods.
2. Measure-Theoretic Controllability
The paper introduces a robust, measure-theoretic definition of controllability. Rather than claiming every single point can reach every other point (a condition easily broken by a single singular point), it proves that the set of reachable points has full Lebesgue measure. For most engineering applications, this "almost everywhere" controllability is sufficient and more resilient to modeling noise.
3. The Spectral Gap
Perhaps the most practical contribution is the use of the spectral gap ($\lambda$) of the sub-Laplacian as a quantitative metric. Instead of a binary "Yes/No" for controllability, the spectral gap defines a "degree of controllability." A larger gap indicates a system that is easier to control, requiring lower $L^2$ norm control effort to reach a target state.
Main Results
- Theorem 1.3: Establishes that if the sub-Laplacian $\Delta_H$ is invertible on the space of zero-mean functions, then for any target ball $B_R(y)$, the set of starting states that can reach that ball has full measure in the domain.
- Corollary 1: Shows that a Poincaré inequality acts as a certificate for controllability. If a constant $\lambda > 0$ exists that satisfies the inequality, the system is certified as approximately controllable.
- Theorem 3.2: Proves that even the weaker condition of operator injectivity is sufficient to guarantee that reachable sets have full measure.
Limitations and Future Directions
While this linear test offers a powerful new tool, several avenues for future research remain:
- Driftless Systems: The current results are limited to control-affine systems without drift. Extending this to systems with natural dynamics (drift) is a necessary next step.
- Computational Discretization: Although the test is linear, the operator is infinite-dimensional. Practical implementation requires discretization (e.g., via basis functions), which may still face the "curse of dimensionality" in very high-dimensional spaces.
- Boundary Regularity: The flow generated by the constructed controls may lack regularity at the boundaries of the domain, potentially affecting the diffeomorphism of the mapping.
By lifting nonlinear reachability into infinite-dimensional functional spaces, this work transforms a difficult geometric problem into a spectral analysis task, providing a new bridge between optimal transport and global control theory.
Summary updated