Implementation and Experimental Validation of Data-Driven Predictive Control for Dissipating Stop-and-Go Waves in Mixed Traffic (2204.03747v3)

Abstract: In this paper, we present the first experimental results of data-driven predictive control for connected and autonomous vehicles (CAVs) in dissipating traffic waves. In particular, we consider a recent strategy of Data-EnablEd Predicted Leading Cruise Control (DeeP-LCC), which bypasses the need of identifying the driving behaviors of surrounding vehicles and directly relies on measurable traffic data to achieve safe and optimal CAV control in mixed traffic. We present the implementation details of DeeP-LCC, including data collection, equilibrium estimation, and control execution. Based on a miniature experiment platform, we reproduce the phenomenon of stop-and-go waves in two typical traffic scenarios: 1) open straight-road scenario under external disturbances and 2) closed ring-road scenario with no bottlenecks. Our experiments clearly demonstrate that DeeP-LCC enables one or a few CAVs to dissipate the traffic waves in both traffic scenarios. These experimental findings validate the great potential of DeeP-LCC in smoothing practical traffic flow in the presence of noisy data, uncertain low-level vehicle dynamics, and communication and computation delays. The code and videos of our experimental results are available at https://github.com/soc-ucsd/DeeP-LCC.

• The paper demonstrates that data-driven DeeP-LCC significantly dampens stop-and-go waves in mixed traffic using real-time vehicular inputs.
• The control method bypasses traditional model identification by employing receding horizon optimization based on measurable speeds and spacing.
• Experimental results in open-road and ring-road scenarios show reduced aggregated squared velocity errors, proving enhanced traffic stability.

Insights into the Implementation and Experimental Validation of Data-Driven Predictive Control in Mixed Traffic

This paper presents the results of an experimental validation of a data-driven predictive control strategy, known as Data-EnablEd Predictive Leading Cruise Control (DeeP-LCC), for mitigating stop-and-go waves in mixed traffic environments involving connected and autonomous vehicles (CAVs) alongside human-driven vehicles (HDVs). The research focuses on bypassing the system identification processes required for traditional model-based controls by directly utilizing input/output traffic data generated in real-time. This approach leverages DeePC, which is rooted in Willems' fundamental lemma, allowing the control of CAVs without explicit models of human-driving behaviors.

Experimental Design and Methodology

Experiment Setup: The experiments utilized a closed environment with miniature robotic vehicles designed to emulate real-world driving dynamics. Two prominent traffic scenarios were considered: an open straight-road scenario affected by external disturbances and a ring-road setting devoid of obvious constraints or bottlenecks, often used to paper intrinsic traffic waves.

Control Strategy: In both scenarios, DeeP-LCC was deployed to control one or a few autonomous vehicles, tasked with dampening traffic waves. The paper detailed the procedural steps in implementing DeeP-LCC, including data collection, predictive control, and control execution. By using only measurable data such as vehicle velocities and spacing, DeeP-LCC generates control strategies without requiring predefined HDV behavior models.

Technical Details:

• Input variables included vehicle accelerations for CAVs and velocity deviations of all vehicles from an equilibrium speed.
• The control problem was framed as an optimization problem constrained by safety (e.g., collision avoidance), equipped with a receding horizon principle to adjust real-time input based on updated conditions.
• Practical implementation challenges like communication delays and real-world data noise were addressed, showcasing the robustness of DeeP-LCC.

Results and Interpretation

Effectiveness in Real Traffic Simulation: The experiments validated that DeeP-LCC control of one or two CAVs considerably mitigated traffic oscillations in mixed traffic, corroborated by time-series data showing reduced velocity fluctuations and more stable traffic flow post-intervention.

Performance Metrics: The effectiveness of DeeP-LCC in suppressing stop-and-go waves was quantitatively assessed using metrics such as aggregated squared velocity errors (ASVE). With CAVs integrated, there were marked reductions in ASVE, indicating lower traffic disturbances and enhanced stability not only at high penetration rates but also at relatively low ones.

Implications and Future Directions

Implications for Traffic Management: The paper indicates significant potential for CAVs, even in mixed traffic environments with low CAV penetration rates, to enhance traffic flow stability and decrease energy consumption due to oscillatory driving patterns.

Adapting to Real-World Challenges: DeeP-LCC’s reliance on real-time data circumvents challenges associated with model inaccuracies and system identification, illustrating how data-driven models can function effectively with the inherent complexities of real-world operations.

Potential Areas for Future Research: Future endeavors could explore adaptive data collection to encompass varying equilibrium conditions more robustly and investigate scaling this approach for urban traffic grids. Additionally, examining the integration with human-driven data collected from real vehicles could offer richer insights and broader applicability in varied traffic scenarios.

This paper contributes to the literature on intelligent transportation systems by offering a fresh perspective on leveraging CAVs for traffic flow optimization within multi-vehicle cooperative environments. The promising results pave the way for adopting data-centric control methodologies in real-world traffic systems.