AutoML-BIMCTS: Optimizing Information Flow Topology for Heterogeneous Vehicle Platoons Under Communication Constraints

The topology of information flow plays a crucial role in controlling connected and automated vehicle platoon. In traditional information flow topologies (IFTs), a uniform communication pattern limits control effectiveness in heterogeneous platoons and under communication constraints. This study aims to enhance the control of heterogeneous vehicles with inferior dynamic performance in the platoon by optimizing information flow topologies within the constraints of communication costs. The design of the information flow topology (IFT) is modeled as an optimization problem with communication cost limitations serving as the constraint. A new hybrid approach, automated machine learning bound improved Monte Carlo tree search (AutoML-BIMCTS), which combines automated machine learning with an enhanced Monte Carlo tree search (MCTS) algorithm, is proposed to address this optimization challenge. Four attention weights, based on position and dynamic performance differences, are integrated with a third-order integral model controller (TIMC) to enhance recognition of the driving status of the inferior vehicle within the platoon. The results demonstrate significant improvements with optimized IFTs, reducing tracking error (TE) by up to 60.51% and fuel consumption by up to 13.44% compared to traditional topologies. PPTE-nps decreases range from 2.93% to 32.25%, and valley position TE drops between 17.19% and 57.43%. This indicates the effectiveness of the optimized topology in preventing separation or rear-end collisions during platoon oscillation. This research contributes a strategic framework for the adaptive control of connected vehicle platoons, offering insights into the balance between communication costs and operational efficiency.