The Turnpike Property in Long Horizon Optimal Control

29 concepts11 papers2 proposals

Imagine planning a road trip from San Francisco to New York. Regardless of your exact starting neighborhood or your precise final destination, the vast majority of your journey will be spent cruising along Interstate 80. You only deviate from this central highway at the very beginning to get on the ramp, and at the very end to exit toward your destination.

In control theory, this phenomenon is known as the turnpike property. When solving an optimal control problem over a long time horizon, the optimal trajectory quickly steers toward a specific steady-state configuration, stays close to it for the bulk of the duration, and only leaves it at the very end to satisfy terminal constraints.

Why the Turnpike Emerges: Dissipativity and Efficiency

Why do completely different initial conditions and objectives yield trajectories that spend most of their time in the exact same state? The answer lies in the fundamental relationship between long-term efficiency and system physics.

A system exhibits a turnpike because the steady state it converges to represents the single most efficient operating point for an infinite-horizon version of the problem. If a finite-horizon trajectory spends time anywhere else, it incurs a continuous cost penalty. As the time horizon grows longer, the cumulative penalty for staying away from this optimal state outweighs the temporary cost required to steer toward it and away from it at the boundaries.

In On turnpike and dissipativity properties of continuous-time optimal control problems, Timm Faulwasser et al. mathematically unified this intuition by establishing the links between turnpike behavior, optimal steady-state operation, and dissipativity. Dissipativity is a generalized notion of energy conservation: a dissipative system absorbs more supply (or incurs more cost) from the outside world when it operates far from its optimal equilibrium. By proving new converse theorems, the authors showed that if an optimal control problem possesses the turnpike property, the underlying system is inherently dissipative with respect to that steady state.

Similarly, in Integral and measure-turnpike properties for infinite-dimensional optimal control systems, Emmanuel Trélat connected this behavior to strict dissipativity and strong duality, formalizing how optimal trajectories are bound to a generalized "turnpike set" rather than just a single point. Maintaining stability and calculating these infinite-horizon targets efficiently can be challenging, prompting methods like the frequency-domain fixed-point iteration algorithm proposed by Joudi Hajar et al. in The Distributionally Robust Infinite-Horizon LQR to guarantee exponential convergence without heavy computational overhead.

From Fixed Points to Periodic Highways

The traditional turnpike property assumes the optimal "highway" is a static, unchanging equilibrium. However, many real-world systems operate best in cycles. For instance, an engine or an economic market might achieve maximum efficiency by repeating a periodic loop rather than sitting still.

To account for this, researchers extended the turnpike framework to periodic orbits. In Periodic optimal control, dissipativity and MPC, Lars Grüne extended the concept of dissipativity to periodic systems, providing a theoretical foundation to guarantee that economic Model Predictive Control (MPC) schemes will naturally converge to these optimal cycles.

When systems grow more complex—such as having time-varying or periodic coefficients—the turnpike property still holds, often with remarkable mathematical precision. For example, in The Exponential Turnpike Property for Periodic Linear Quadratic Optimal Control Problems in Infinite Dimension, Emmanuel Trélat et al. proved an exponential periodic turnpike property for infinite-dimensional linear-quadratic problems. They demonstrated that over long horizons, the optimal trajectory doesn't just drift toward a periodic orbit; it converges to it at an exponential rate.

Scaling Up to Infinite Dimensions and Mean Fields

One of the most powerful aspects of the turnpike property is its ability to simplify incredibly complex, high-dimensional problems. As reviewed by Borjan Geshkovski in Turnpike in optimal control of PDEs, ResNets, and beyond, time-dependent control problems governed by partial differential equations (PDEs) or deep learning architectures like Residual Networks (ResNets) tend to track the solutions of their corresponding time-independent, static counterparts. This means engineers can approximate a grueling, time-dependent dynamic optimization problem by solving a much simpler static one.

This simplification scales across various mathematical abstractions:

Practical Applications: Robotics and Supply Chains

Beyond pure theory, identifying turnpikes unlocks massive computational advantages in practical engineering.

In robot motion planning, calculating an optimal path through a complex, high-dimensional configuration space usually requires expensive, redundant optimization across the entire trajectory. However, in Geodesic turnpikes for robot motion planning, Yann de Mont-Marin et al. formalized "geodesic turnpikes" for robots, showing that optimal paths consist of long steady-state motion segments flanked by short transitions. By identifying these turnpikes as saddle points in Lie algebra dynamics, the authors achieved an order-of-magnitude speedup in motion planning solvers.

Similarly, in logistics and supply chain management, Sophie Hall et al. leveraged the turnpike property within receding horizon games in Receding Horizon Games for Modeling Competitive Supply Chains⋆. Because the open-loop trajectories naturally converge toward an equilibrium state before diverging near the horizon's end, the framework can tractably simulate how competitive manufacturers make decisions under uncertainty without collapsing under computational weight.

Ultimately, the turnpike property works because nature and economics favor efficiency. Over a long enough timeline, the transient costs of getting onto and off the optimal highway become negligible, making the steady state the mathematical gravity well of optimal control.

Key Foundational Work

On turnpike and dissipativity properties of continuous-time optimal control problems (2017) — contributed 3 concept elements to this direction.
Periodic optimal control, dissipativity and MPC (2014) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_turnpike_property_64249adc,
  title = {turnpike property},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/turnpike-property/rdr_64249adc7a5e483e8f6da0d3f04d0361},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_turnpike_property_64249adc_v1,
  title = {turnpike property},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/turnpike-property/rdr_64249adc7a5e483e8f6da0d3f04d0361/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake