In safety-critical control, the goal is to ensure a system never enters a "forbidden" region of its state space. Modern frameworks like Control Barrier Functions (CBFs) achieve this by solving an optimization problem—typically a Quadratic Program (QP)—at every time step. While these methods provide rigorous mathematical guarantees of safety, they often fail a crucial test of physical reality: the resulting control signals can be jerky, abrupt, or "chattering."
This research direction focuses on regularity—ensuring that control laws are Lipschitz continuous. A Lipschitz continuous controller ensures that small changes in the system's state lead to only small changes in the control input, preventing the mechanical wear and instability that plague discontinuous "bang-bang" controllers.
The Problem of Actuator Chattering
When a robot or power grid operates near the boundary of safety, optimization-based controllers often switch abruptly between different constraints. As Tabuada et al. note, these discontinuous jumps are incompatible with the physical limitations of real-world actuators. In power networks, for instance, Cortes et al. explain that abrupt interventions cause mechanical stress and high-frequency oscillations, which can damage hardware even if the system remains "safe" in a purely mathematical sense.
Beyond hardware wear, discontinuity creates a theoretical crisis. Standard proofs for the existence and uniqueness of a system's trajectory rely on the control law being locally Lipschitz continuous. If the controller is discontinuous, the system may exhibit unpredictable behavior or fail to satisfy the very safety guarantees it was designed to uphold.
Engineering Smoothness into Optimization
To fix this, researchers have developed methods to bake regularity directly into the control design. The challenge is to maintain safety without losing the computational efficiency of optimization.
KKT Conditions and Closed-Form Solutions
One approach is to derive the "answer" to the optimization problem analytically. Tabuada et al. demonstrated that by using Karush–Kuhn–Tucker (KKT) optimality conditions, one can prove that a CBF-based controller remains locally Lipschitz continuous as long as the control input maintains a direct influence on the safety barrier. Building on this, Wang et al. derived explicit solutions for safety problems, allowing low-power embedded devices to look up smooth control signals in a pre-computed map rather than solving complex optimizations in real-time.
Reformulating the Problem Structure
Sometimes the structure of the optimization problem itself is the culprit. Standard QPs can become discontinuous when their constraints change rapidly. Agrawal et al. proposed a novel fix: reformulating QPs into Second-Order Cone Problems (SOCPs). This transformation guarantees Lipschitz continuity regardless of the specific constraint structure, bypassing the need for complex checks like the linear independence of constraint gradients.
Regularity Under Uncertainty
Real-world systems rarely have perfect models. When we add robustness to the mix—designing controllers that work despite measurement errors or noise—maintaining smoothness becomes even harder.
Mestres et al. established that if safety and stability constraints are strictly compatible, the resulting robust control signals are guaranteed to be locally Lipschitz. This is vital for "distributionally robust" controllers, which use data samples to handle uncertainty. Long et al. showed that these data-driven controllers exhibit "point-Lipschitz" continuity, ensuring that even when the robot is learning from noisy data, its movements remain fluid and safe for its mechanical joints.
Bridging the Gap Between Theory and Hardware
The shift toward regularized control laws allows researchers to bridge disparate fields. For example, Delimpaltadakis et al. showed that smooth CBF-based systems can be used to approximate "Projected Dynamical Systems"—discontinuous systems often used in optimization—allowing for easier numerical simulation and more robust implementation.
Similarly, in the realm of reachability analysis, Borquez et al. and the authors developed filters that move away from "jittery" bang-bang behavior toward hardware-friendly, optimization-based trajectories. By prioritizing Lipschitz continuity, these researchers ensure that the mathematical elegance of safety certificates translates into the reliable, smooth operation of physical machines.
Key Foundational Work
Input-to-state stabilization of feasible model predictive controllers for linear systems (2011) — contributed 1 concept elements to this direction.
Piecewise-affine state feedback for piecewise-affine slab systems using convex optimization (2005) — contributed 1 concept elements to this direction.