Cooperative Longitudinal Control Of Intelligent And Connected Vehicles In Mixed Traffic Flow

As the deployment of Intelligent and Connected Vehicles（ICVs）,modern transporta-tion systems are going to face a long-period mixed traffic phase,where ICVs and Human-Driven Vehicles（HDVs）coexist.Cooperative control is one of the key technologies of ICVs,but existing research mostly focuses on the full-ICV scenarios,which is hardly ap-plicable to the mixed traffic environment.Meanwhile,due to the unknown behavior of HDVs,model-based control methods are also inapplicable to mixed traffic.To address these problems,this paper presents the influence mechanism of ICVs for mixed traffic,proposes the notion and design method of Leading Cruise Control（LCC）in mixed traffic,and further proposes a model-free data-driven predictive control method for LCC,provid-ing a theoretical basis and supporting methodology for cooperative longitudinal control of ICVs in mixed traffic flow.Firstly,through the analysis of controllability and formation pattern,the inner mech-anism is revealed for ICVs in stabilizing and improving mixed traffic flow.The research results point out:1)In the case of homogeneous or heterogeneous driving behavior of HDVs,the mixed traffic system can be stabilized under the influence of single or multi-ple ICVs.2)The non-cooperative formation of ICVs has some limitations with respect to stability invarience property and diminishing marginal improvement property ofH₂performance.3)The cooperative formation of ICVs has two kinds of optimal patterns:uniform distribution and platoon formation.When the other vehicles have a car-following behavior with poor stability,the uniform distribution has the largest potential in improving mixed traffic performance.Secondly,by making full use of the internal mechanism of ICV’s influence on mixed traffic,the notion and design method of LCC is proposed,which integrates the following and leading functions of ICVs,is applicable under general communication topologies,and actively improves the mixed traffic performance.The research results point out:1)From the perspective of leading the vehicles behind,controllability and observability re-sults reveal the feasibility of LCC in improving upstream traffic performance.2)From the perspective of following the vehicles ahead,the head-to-tail string stability results re-veal the superiority of LCCs in mitigating downstream traffic perturbations.3)Based on structuredH₂optimal control,LCC is able to resolve the problem of ICV optimal control in mixed traffic under communication constraints.Thirdly,based on LCC,a model-free data-driven predictive control method,DeeP-LCC,is developed.The research results point out:1)Under the centralized control ar-chitecture,the method uses Willems’fundamental lemma to construct the data-driven dynamics representation of mixed traffic.It employs the safety-constrained predictive control framework to centrally solves the control input,resolving the model-free,safe and optimal cooperative control problem of ICVs in mixed traffic.2)Under the dis-tributed control architecture,based on the local trajectory data of the subsystem and the coupling dynamic relationship between the subsystems,this method uses the alternat-ing direction method of multipliers to design a distributed optimization algorithm.This method resolves the model-free,real-time and efficient cooperative control problem of ICVs in large-scale mixed traffic flow.Finally,a cloud control miniature test platform is built for real-vehicle tests of DeeP-LCC.The research results point out:in the case of multi-scenario,multi-penetration-rate and multi-formation configurations,DeeP-LCC is able to stabilize and improve the mixed traffic flow under the influence of real-world traffic factors including noise disturbance,communication and computing delays,and integrated cloud-vehicle dynamics.