In traditional control theory, the goal is usually to make a system follow a specific path or stay at a fixed setpoint. We design a "tracking" cost—typically the distance from the target—which naturally acts as a guide for the system. However, in many real-world applications like power grids, chemical plants, or wind turbines, we don't just want to stay at a point; we want to maximize profit or minimize energy consumption.
This shift to Economic Model Predictive Control (EMPC) introduces a fundamental problem: economic costs (like "maximize power") do not look like tracking costs. They don't have a natural minimum at the desired operating point, and they don't inherently guarantee stability. A controller might "greedily" deplete the system's stored energy to get a short-term economic boost, leading to a crash later.
The research direction emerging from these papers centers on two deeply linked concepts—Dissipativity and the Turnpike Property—which provide the mathematical "missing links" to ensure that economic optimization leads to stable, reliable behavior.
The Turnpike Property: The Geometry of Long-Term Planning
If you are driving from New York to Los Angeles, the most efficient route isn't a straight line; it’s getting on the highway (the "turnpike"), staying there for the vast majority of the trip, and only getting off near your destination.
Martin Gugat and others have shown that optimal control trajectories behave exactly like this. In Modeling, Control and Numerics of Gas Networks, Gugat explains that for long-horizon problems, the system spends most of its time near a specific "steady state" equilibrium, regardless of where it started or where it needs to end.
This isn't just a curiosity; it’s a powerful simplification tool. As Yizhou Zhou demonstrates in Exponential turnpike property for particle systems and mean-field limit, this property holds even for massive systems of interacting particles or robots. It tells us that for long-duration tasks, we can approximate complex dynamic planning by simply finding the best static operating point.
Dissipativity: The Engine Behind the Turnpike
Why do optimal paths gravitate toward these "turnpikes"? The answer lies in Dissipativity.
A system is dissipative if it "leaks" a generalized version of energy toward an optimal state. Matthias A. Müller argues in Dissipativity in economic model predictive control: beyond steady-state optimality that dissipativity is the unifying theory for EMPC. It provides a "storage function"—a mathematical way to account for the energy stored in the system—that bridges the gap between economic performance and stability.
Timm Faulwasser and Milan Korda proved in On turnpike and dissipativity properties of continuous-time optimal control problems that these two ideas are two sides of the same coin: if a system exhibits turnpike behavior, it must be dissipative, and vice versa. This equivalence allows engineers to use the geometric behavior of a system to prove its underlying stability.
Cost Rotation: Making Economics Behave Like Tracking
The most practical breakthrough in this field is the technique of Cost Rotation. Since economic costs don't have a minimum at the target, we "rotate" them using Lagrange multipliers from an offline optimization.
Mario Zanon and Moritz Diehl pioneered this in A Lyapunov function for periodic economic optimizing Model Predictive Control. By adding a storage function to the economic cost, they create a "rotated cost" that does have a minimum at the optimal trajectory. This transformed cost acts as a Lyapunov function, a mathematical certificate that the system will safely converge to the target.
This rotation allows us to treat "profit maximization" with the same mathematical rigor as "staying at a setpoint." It prevents the "greedy" behavior seen in naive controllers, such as the wind turbine deceleration problem described by Sebastien Gros in An economic NMPC formulation for wind turbine control.
Generalizing to Periodic Orbits and Stochastic Noise
Modern research is pushing these boundaries beyond simple steady states. Often, the most "economic" way to run a system is a cycle (like a walking gait or a chemical batch process).
Lukas Schwenkel and Matthias A. Müller have developed "discounted" EMPC methods that stabilize these periodic orbits without needing to know the cycle's exact shape beforehand. In Linearly discounted economic MPC without terminal conditions for periodic optimal operation, they show that simply fading out the cost over the horizon can force the system to settle into its most efficient rhythm.
Finally, Mario Zanon has extended this entire framework to Markov Decision Processes (MDPs). By lifting dissipativity from individual states to the space of probability distributions, researchers can now provide stability guarantees for economic controllers operating in highly noisy, stochastic environments.
By combining the geometric intuition of the turnpike with the energetic rigor of dissipativity, this field has turned "economic control" from a heuristic-heavy practice into a precise science.