Distributed Cooperative MPC for Autonomous Driving in Different Traffic Scenarios

Mohseni, Fatemeh; Frisk, Erik; Nielsen, Lars · 2020 · OpenAlex-citations

DOI: 10.1109/tiv.2020.3025484

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of controlling multiple autonomous vehicles in complex traffic scenarios, such as double lane-switching and intersection maneuvers, using a Distributed Cooperative Model Predictive Control (DCMPC) approach. The primary motivation is to overcome the computational limitations of centralized control methods, which struggle with scalability as the number of vehicles increases. By distributing the control problem, the authors aim to achieve real-time computational efficiency while ensuring safety through collision avoidance and maintaining system stability. The work specifically focuses on handling non-holonomic vehicle dynamics and uncertainties arising from the deviation between estimated and actual trajectories of neighboring vehicles. The methodology formulates the control task as a distributed optimal control problem for a system of $n$ autonomous vehicles. Each vehicle is modeled using kinematic equations that account for position, orientation, speed, and steering angle, subject to physical constraints like speed limits and steering rates. To manage computational complexity, the centralized problem is decomposed into individual optimization problems for each vehicle. A key innovation is the introduction of a "compatibility constraint" alongside standard collision avoidance constraints. This compatibility constraint limits the deviation of a vehicle’s actual trajectory from the trajectory estimated by its neighbors, thereby accounting for uncertainty and ensuring that vehicles do not deviate significantly from expected paths. Each vehicle computes its control inputs based on estimated states of neighbors, utilizing a terminal controller to guarantee stability within a defined terminal region. The study establishes theoretical guarantees for the proposed DCMPC algorithm, proving asymptotic convergence of the vehicle system to the desired destination in the absence of disturbances, provided the MPC control horizon is sufficiently small. The authors demonstrate that the compatibility constraint, combined with the predictive collision avoidance feature, ensures both safety and convergence. Simulation results indicate that the distributed algorithm scales effectively with an increasing number of vehicles, offering significant computational gains compared to centralized approaches. The method successfully handles various traffic scenarios, including interactions with non-autonomous vehicles, by robustly managing trajectory uncertainties and maintaining safe distances. The significance of this work lies in its contribution to scalable and safe cooperative control for autonomous driving systems. By providing a general framework applicable to different traffic maneuvers, the DCMPC approach reduces reliance on centralized infrastructure, enhancing the autonomy of individual vehicles. The integration of compatibility constraints addresses a critical gap in existing distributed MPC methods by explicitly handling trajectory estimation errors, thus improving robustness. This research supports the development of efficient, real-time control strategies for multi-agent autonomous systems, facilitating smoother traffic flow and reduced congestion through cooperative vehicle behavior.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness