Ramp Merging Strategy Based on Dual-layer Optimization Model in Non-connected Scenarios

The ramp merging task is a major challenge in the development of autonomous driving technology. In scenarios where all other vehicles in the merging area are manually driven, the ego driving vehicle must rely on its own sensors and communication with the roadside unit to obtain environmental information, make decisions, and complete the merging task in a non-cooperative setting. However, the current research on ramp merging lacks the capability to dynamically adjust merging gaps and longitudinal speeds, and also exhibits limited interaction intelligence with mainline vehicles. To address this problem, this paper introduces a staged maneuvering model based on optimal control. Simultaneously, a merging strategy grounded on the dual-layer optimization model predictive control is proposed, empowering ego driving vehicles to make more intelligent and efficient merging decisions. Firstly, a python-based simulation platform is established to verify the ramp merge decision. After that, a staged maneuvering model is developed based on the optimal control method, and both the proposed merging strategy based on the dual-layer optimization model predictive control and the rule-based baseline control strategy are introduced. The merging process is then simulated under two scenarios using the baseline strategy and the proposed strategy, respectively. The results are analyzed, demonstrating that the proposed algorithm enables safer and smarter decisions. The proposed strategy achieves dynamic selection of the merging gap, thereby enhancing the intelligence of the ramp merging process, achieving both efficiency and safety, and providing a reliable method for vehicle ramp merging in non-cooperative environments.

I. Introduction

The potential of autonomous driving technology to enhance traffic efficiency and reduce traffic incidents has been recognized [1]. Additionally, the ramp merging task stands as one of the critical challenges in the advancement of autonomous driving technology. The main difficulties lie in the game interaction and collaboration between the vehicles entering the ramp and those on the mainline, as well as predicting and adapting to the intentions and behaviors of the mainline vehicles [2]. Ramp merging on highways is a high-risk maneuver that is prone to traffic accidents and conflicts due to varying driving styles, differences in traffic conditions between the mainline and ramp, and road design factors [3]. In non-cooperative scenarios, the process of merging vehicles from the ramp to the mainline involves two critical components: longitudinal speed adjustment in the acceleration lane, and lane change to the desired gap in the mainline lane. These two processes are dynamically and interactively coupled. However, the pass ability of the merging gap may dynamically change during the actual merging process, necessitating the autonomous vehicle to dynamically select the merging gap and adjust its speed accordingly. The optimization-based approach is a crucial method for ramp merging decision-making, enabling the attainment of optimal solutions in diverse environments through various modeling methods, effectively integrating both decision-making and planning aspects [4] [5].