Novel Decision-Making Strategy for Connected and Autonomous Vehicles in Highway On-Ramp Merging

Abstract

High-speed highway on-ramp merging is a significant challenge toward realizing fully automated driving ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">level 4</i> ). Connected Autonomous Vehicles ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CAVs</i> ), that combine communication and autonomous driving technologies, may improve greatly the safety performances when performing highway on-ramp merging. However, even with the emergence of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CAVs</i> , some keys constraints should be considered to achieve a safe on-ramp merging. First, human-driven vehicles will still be present on the road, and it may take decades before all the commercialized vehicles will be fully autonomous and connected. Also, onboard vehicle sensors may provide inaccurate or incomplete data due to sensors limitations and blind spots, especially in such critical situations. To resolve these issues, the present work introduces a novel solution that uses an off-board Road-Side Unit ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RSU</i> ) to realize fully automated highway on-ramp merging for connected and automated vehicles. Our proposed approach is based on an Artificial Neural Network (ANN) to predict drivers’ intentions. This prediction is used as an input state to a Deep Reinforcement Learning ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DRL</i> ) agent that outputs the longitudinal acceleration for the merging vehicle. To achieve this, we first propose a data-driven model that can predict the behavior of the human-driven vehicles in the main highway lane, with <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">99</i> % accuracy. We use the output of this model as input state to train a Twin Delayed Deep Deterministic Policy Gradients ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TD3</i> ) agent that learns “ <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">safe</i> ” and “ <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">cooperative</i> ” driving policy to perform highway on-ramp merging. We show that our proposed decision-making strategy improves performance compared to the solutions proposed previously.