[Paper Review] Distributed Nonlinear Model Predictive Control and Metric Learning for Heterogeneous Vehicle Platooning with Cut-in/Cut-out Maneuvers

Mohammad Hossein Basiri, Benyamin Ghojogh+4 more|arXiv (Cornell University)|Apr 1, 2020

Traffic control and managementEngineering28 references33 citations

TL;DR

This paper proposes a Distributed Nonlinear Model Predictive Control (DNMPC) framework for heterogeneous vehicle platooning that handles cut-in and cut-out maneuvers while ensuring collision avoidance, safe spacing, and desired speed tracking. By integrating distributed metric learning and ADMM optimization, the method enhances driving comfort, fuel economy, and convergence performance, with simulation results confirming stability and effectiveness across multiple unidirectional topologies including TPLF and PLF.

ABSTRACT

Vehicle platooning has been shown to be quite fruitful in the transportation industry to enhance fuel economy, road throughput, and driving comfort. Model Predictive Control (MPC) is widely used in literature for platoon control to achieve certain objectives, such as safely reducing the distance among consecutive vehicles while following the leader vehicle. In this paper, we propose a Distributed Nonlinear MPC (DNMPC), based upon an existing approach, to control a heterogeneous dynamic platoon with unidirectional topologies, handling possible cut-in/cut-out maneuvers. The introduced method addresses a collision-free driving experience while tracking the desired speed profile and maintaining a safe desired gap among the vehicles. The time of convergence in the dynamic platooning is derived based on the time of cut-in and/or cut-out maneuvers. In addition, we analyze the improvement level of driving comfort, fuel economy, and absolute and relative convergence of the method by using distributed metric learning and distributed optimization with Alternating Direction Method of Multipliers (ADMM). Simulation results on a dynamic platoon with cut-in and cut-out maneuvers and with different unidirectional topologies show the effectiveness of the introduced method.

Motivation & Objective

• To address the challenge of dynamic vehicle platooning with heterogeneous vehicles under cut-in and cut-out maneuvers.
• To ensure collision-free operation while maintaining safe inter-vehicle gaps and tracking desired speed profiles.
• To improve driving comfort, fuel economy, and convergence performance through distributed optimization and metric learning.
• To derive a theoretical convergence time for the DNMPC system based on maneuver timing.
• To validate the method across multiple unidirectional communication topologies (e.g., TPLF, PLF, TPF, PF).

Proposed method

• A distributed nonlinear MPC (DNMPC) framework is extended from a prior approach [15] to handle dynamic platooning with heterogeneous vehicles.
• The control problem is formulated as a constrained optimization over a predictive horizon, minimizing a cost function that balances tracking, control effort, and safety.
• Distributed metric learning is applied to learn subspace metrics for driving comfort (F), fuel economy (R), and convergence (Q, G) using ADMM optimization.
• ADMM is used to solve the distributed optimization problem in a scalable and computationally efficient manner.
• The method incorporates vehicle-specific dynamics, including mass, drag, rolling resistance, and driveline efficiency, in a nonlinear state-space model.
• The system uses predicted and assumed outputs to handle communication delays and uncertainty in real-time control.

Experimental results

Research questions

• RQ1 How can a distributed nonlinear MPC framework be extended to handle cut-in and cut-out maneuvers in heterogeneous vehicle platoons?
• RQ2 What is the theoretical convergence time of the DNMPC system in dynamic platooning, and how does it depend on maneuver timing?
• RQ3 To what extent does distributed metric learning improve driving comfort, fuel economy, and convergence in platooning control?
• RQ4 How do different unidirectional communication topologies (e.g., TPLF, PLF) affect the performance of the DNMPC system?
• RQ5 Can ADMM-based optimization effectively support real-time distributed control with metric learning in dynamic platooning?

Key findings

• The DNMPC system achieved convergence within 11 seconds after the last maneuver (cut-out at t=4s), consistent with the theoretical convergence time derived in Theorem 2.
• Driving comfort was significantly improved over time, with predicted and assumed outputs converging in the metric subspace, especially after transient disturbances from cut-in and cut-out maneuvers.
• Fuel economy was enhanced by minimizing abrupt changes in control inputs, as shown by sparse and smooth projected control input trajectories in the R-weighted metric subspace.
• Absolute and relative convergence were improved, with predicted outputs closely tracking desired outputs and neighboring vehicle states, particularly after 7 seconds of algorithm progression.
• The TPLF and PLF topologies showed superior performance in maintaining stable spacing and velocity profiles, with minimal oscillations in relative spacing error.
• The use of distributed metric learning with ADMM optimization enabled real-time adaptation of performance metrics, leading to measurable improvements in comfort, fuel efficiency, and convergence stability.