Decentralized Robust Data-driven Predictive Control for Smoothing Mixed Traffic Flow

TL;DR

This paper proposes a decentralized robust DeeP-LCC (Data-EnablEd Predictive Leading Cruise Control) approach for CAVs to smooth mixed traffic, which greatly reduces computation complexity with better safety performance, while naturally preserving data privacy.

Abstract

In a mixed traffic with connected automated vehicles (CAVs) and human-driven vehicles (HDVs) coexisting, data-driven predictive control of CAVs promises system-wide traffic performance improvements. Yet, most existing approaches focus on a centralized setup, which is not computationally scalable while failing to protect data privacy. The robustness against unknown disturbances has not been well addressed either, causing safety concerns. In this paper, we propose a decentralized robust DeeP-LCC (Data-EnablEd Predictive Leading Cruise Control) approach for CAVs to smooth mixed traffic flow. In particular, each CAV computes its control input based on locally available data from its involved subsystem. Meanwhile, the interaction between neighboring subsystems is modeled as a bounded disturbance, for which appropriate estimation methods are proposed. Then, we formulate a robust optimization problem and present its tractable computational solutions. Compared with the centralized formulation, our method greatly reduces computation burden with better safety performance, while naturally preserving data privacy. Extensive traffic simulations validate its wave-dampening ability, safety performance, and computational benefits.

Figures (7)

• Figure 1: Schematic of centralized and decentralized robsut DeeP-LCC for CAVs in mixed traffic. (a) Centralized DeeP-LCC. It collects the data of the whole mixed traffic system, including velocity errors of all vehicles (black dashed arrow) and spacing errors of CAVs collectively (blue dashed arrow), and utilizes these data to design control input of CAVs (red squiggle arrow). (b) Decentralized robust DeeP-LCC. It decomposes the mixed traffic system into multiple CF-LCC subsystems and only requires locally available data. The dynamic coupling between subsystems is modeled as the disturbance (purple squiggle arrow).
• Figure 2: Schematic of comparison between centralized and decentralized robust DeeP-LCC. The Hankel matrix is partitioned into the past trajectories (represented by blue columns) and future trajectories (represented by red columns). The future external disturbance is represented by purple squiggle arrows.
• Figure 3: Schematic of three disturbance estimation methods. The purple line denotes the actual disturbance trajectory, whose past is known while its future needs to be estimated. The zero estimation is denoted as the black dashed line, while the time-varying bound estimated set and the constant bound estimated set are represented as the red region and the blue region, respectively.
• Figure 4: Comparison of computation time for different methods.
• Figure 5: Velocity profiles in Experiment A where the head vehicle is under sinusoidal perturbation. The black profile represents the head vehicle and the gray profile represents the HDVs. The red profile (vehicle 3), blue profile (vehicle 6), orange profile (vehicle 10) and green profile (vehicle 13) represent the first to the fourth CAV, respectively. (a) All vehicles are HDVs. (b) CAVs utilize cDeeP-LCC. (c) CAVs utilize dDeeP-LCC.

Theorems & Definitions (11)

• Definition 1: Persistently Exciting
• Proposition 1: wang2023deep
• Proposition 2: wang2022distributed
• Remark 1: Coupling dynamics and estimation of future velocity errors
• Remark 2: Centralized vs decentralized formulations
• Remark 3: Data requirement in dDeeP-LCC
• Remark 4
• Proposition 3: Method I: Vertex-based Strategy
• Proposition 4: Method II: Duality-based strategy
• Remark 5: Comparison with model-based methods