Vessel-following model for inland waterways based on deep reinforcement learning

With the growth of traffic on inland waterways, autonomous driving technologies for vessels will gain increasing significance to ensure flow and safety. Inspired by car-following models road traffic, which demonstrated their strength reduce stop-and-go waves increase efficiency safety, we propose a vessel-following model waterways based deep reinforcement learning (RL). Our is trained under consideration realistic vessel dynamics environmental influences, such as varying stream velocity river profile, with reward function favoring observed following behavior comfort. Aiming at high generalization capabilities, training environment that uses stochastic processes leading trajectory dynamics. safe comfortable in different unseen scenarios, including Middle Rhine. In comparison an existing model, our was able early anticipate safety–critical situations, resulting higher safety while maintaining comparable further experiments, proposed approach its potential dampen oscillations congestion using sequence followers.