⚡ electrical engineering

Optimal Finite-Horizon LQR Control for Traffic Flow via Variable Speed Limits

This paper proposes a finite-horizon Linear Quadratic Regulator (LQR) framework for controlling traffic flow via variable speed limits, which solves a nonlinear Riccati partial differential equation analytically to achieve superior, time-sensitive performance and guaranteed finite-time convergence compared to traditional infinite-horizon approaches.

Original authors: Brian Block, Stephanie Stockar

Published 2026-06-03

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Brian Block, Stephanie Stockar

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a highway as a giant, flowing river of cars. Sometimes the water is fast and clear (free-flow traffic), and sometimes it gets clogged and slow (congested traffic). The goal of this paper is to figure out how to use "Variable Speed Limits" (VSL)—those electronic signs that tell drivers to slow down or speed up—to smooth out the traffic jams and get everyone moving efficiently.

Here is a simple breakdown of what the authors, Brian Block and Stephanie Stockar, did:

1. The Problem: Traffic is a Wave, Not a Line

The authors treat traffic like a wave in a fluid. They use a mathematical model called the LWR model (named after three scientists) to describe how traffic density changes over time and space.

The Old Way: Previous studies tried to control this "river" using an Infinite Horizon approach. Think of this like trying to steer a ship by looking at the horizon forever, hoping you eventually get to the right spot. It works well for long-term stability, but it doesn't guarantee you'll fix a problem right now.
The New Way: This paper introduces a Finite Horizon approach. This is like setting a timer: "I need to clear this traffic jam in exactly 5 minutes." The controller knows it has a deadline and acts accordingly.

2. The Solution: The "Smart Speed Limit" Brain

The authors developed a new "brain" for traffic lights and speed limit signs. This brain uses a method called LQR (Linear Quadratic Regulator).

How it works: Imagine you are driving a car and you want to reach a specific speed at a specific time. You have to decide: "Do I slam on the brakes hard (which is uncomfortable and uses a lot of energy), or do I gently ease off the gas?"
The Math Magic: To make this decision perfectly, the authors had to solve a very complex mathematical puzzle called a Riccati Equation.
- In the past, this puzzle was solved for a "forever" timeline, resulting in a static answer (like a fixed rulebook).
- In this paper, they solved it for a "countdown" timeline. This means the answer changes not just based on where you are on the road, but also when you are there. It's like a GPS that updates your route every second based on how much time is left to get to your destination.

3. Two Lanes of Traffic

Traffic behaves differently when it's moving fast versus when it's stuck.

Free Flow: Cars are moving freely.
Congested Flow: Cars are bumper-to-bumper.
The authors created a "switching" system. The controller recognizes which "lane" of behavior the traffic is in and applies a different set of rules for each. It's like a coach who gives different instructions to a sprinter (free flow) than to a marathon runner (congested flow).

4. The "Tightrope" of Control

The paper also tested how sensitive this system is to its settings (called parameters Q, R, and S).

The Analogy: Imagine you are trying to balance a broom on your hand.
- If you are too aggressive (trying to fix the jam too fast), you might overcorrect and make the situation worse, or demand speed limits that are impossible (like telling cars to drive 100 mph when the limit is 65).
- If you are too lazy (not caring enough about the jam), the traffic stays stuck.
The Finding: The authors found that if you don't tune these settings carefully, the computer might suggest speed limits that are physically impossible or unsafe. They showed that for short deadlines (Finite Horizon), the controller needs to be very precise to avoid these "impossible" commands.

5. The Results: Faster Fixes

They tested their new "Finite Horizon" controller on two scenarios:

A Circular Road: Like a race track where cars loop around.
A Straight Road: Where cars enter and exit, like a real highway.

The Verdict:
The new controller was much better at clearing traffic jams within a specific time limit compared to the old "forever" controller.

On the straight road, the old controller struggled to handle cars entering the road because it didn't have a deadline to work toward.
The new controller, knowing it had to finish the job in, say, 50 seconds, acted more decisively to clear the congestion before the time ran out.

Summary

In short, this paper teaches us how to build a traffic controller that doesn't just "hope" traffic gets better eventually, but actively plans a step-by-step strategy to clear a jam within a specific time limit. It uses advanced math to figure out the perfect speed limit to display at every moment, ensuring traffic flows smoothly without asking drivers to do the impossible.

Technical Summary: Optimal Finite-Horizon LQR Control for Traffic Flow via Variable Speed Limits

Problem Statement
Traffic congestion remains a critical challenge in urban and highway environments, necessitating controllers capable of alleviating jams within specific time frames. While macroscopic traffic models, such as the Lighthill-Whitham-Richards (LWR) model, are widely used to describe traffic dynamics via partial differential equations (PDEs), existing optimal control strategies often rely on infinite-horizon formulations. These infinite-horizon approaches, while effective for steady-state regulation, fail to guarantee that control objectives are met within a finite time window. Furthermore, prior optimal control applications to the LWR model have largely been restricted to boundary control (e.g., ramp metering) or infinite-horizon in-domain control that operates only within the free-flow regime. There is a lack of frameworks that address finite-time convergence for in-domain control using Variable Speed Limits (VSL) across both free-flow and congested traffic regimes.

Methodology
The authors propose a finite-horizon Linear Quadratic Regulator (LQR) framework applied directly to the first-order hyperbolic LWR traffic model with a triangular fundamental diagram. The control action is realized through Variable Speed Limits (VSL), implemented as a source term in the PDE by modifying the flow-density relationship.

The methodology proceeds through the following steps:

Model Linearization: The nonlinear LWR model is linearized around a nominal equilibrium density and speed ratio. Two distinct linearized models are derived to handle the free-flow ( $\Phi_f$ ) and congested ( $\Phi_c$ ) regimes separately, as the system dynamics differ significantly between these states.
Control Formulation: The control problem is formulated as a finite-horizon optimal control problem. Unlike previous infinite-horizon approaches where the state feedback gain is time-invariant, this formulation introduces time dependence. The cost function includes penalties on state deviations ( $Q$ ), control effort ( $R$ ), and a terminal state weight ( $S$ ) at a final time $t_f$ .
Riccati PDE Solution: The optimal control law is derived by solving an operator Riccati equation. In the finite-horizon case, this equation becomes a nonlinear Partial Differential Equation (PDE) involving both time and space derivatives. The authors solve this Riccati PDE analytically using the parametric method of characteristics.
Feedback Synthesis: The solution to the Riccati PDE yields a state feedback function, $P(z,t)$ , which depends on space, time, and the current traffic regime. The optimal VSL control input is then generated as a switching function based on this feedback gain.

Key Contributions
The paper outlines three primary contributions:

Regime Handling: The extension of a previously developed VSL controller to handle both free-flow and congested traffic regimes. This is achieved by defining two separate linear operators, each valid within its respective regime.
Finite-Horizon Extension: The development of a finite-time optimal control framework for PDEs. This involves deriving an analytical solution to the time-dependent Riccati PDE using the method of characteristics. The resulting state feedback function is a dynamic switching function dependent on time, space, and traffic regime, enabling finite-time convergence.
Parameter Sensitivity Analysis: A comprehensive analysis of the control parameters ( $Q$ , $R$ , $S$ , and $t_f$ ). The study demonstrates that without careful tuning, the control input (speed limit) can become infeasible (exceeding physical limits). The analysis provides guidelines for parameter selection to balance control aggressiveness with feasibility.

Results
The proposed controller was evaluated through simulations on two scenarios: a circular road with periodic boundary conditions and a straight road with time-varying inflow.

Performance Comparison: The finite-horizon controller was shown to guarantee convergence to the desired density profile within the specified time frame, a capability the infinite-horizon controller lacks.
Parameter Sensitivity (Infinite Horizon): Simulations on the circular road indicated that high penalties on state deviation ( $Q$ ) relative to control effort ( $R$ ) lead to aggressive control actions, potentially resulting in infeasible speed limits (e.g., exceeding 85 km/h). Conversely, relaxing the control penalty ( $R$ ) also increased the required speed limits.
Parameter Sensitivity (Finite Horizon): Reducing the final time $t_f$ required more aggressive control actions to meet the deadline, often pushing speed limits to infeasible levels. The terminal weight $S$ had a significant impact on the circular road scenario but a less pronounced effect on the straight road scenario with varying inflow.
Straight Road Dynamics: In the straight road scenario, the finite-horizon controller demonstrated superior ability to mitigate the impact of incoming traffic flows compared to the infinite-horizon approach, particularly when the control horizon was shortened. However, the controller's inability to account for unknown future inflows remained a limitation in certain high-density scenarios.

Significance
The paper claims that the proposed finite-horizon LQR approach offers significantly improved control performance across various traffic scenarios compared to infinite-horizon formulations. By providing a mechanism to guarantee convergence within a specific time window, the method is particularly suitable for time-sensitive traffic management applications. The analytical derivation of the Riccati PDE solution allows for the implementation of a state-feedback controller that adapts to both time and space, addressing a gap in the control of hyperbolic PDEs. The sensitivity analysis underscores the necessity of careful parameter tuning to ensure physical feasibility, laying the groundwork for future work on constrained optimal control.