Applying Decision Forest for Improving the Adaptation of the ML Algorithm for Beamsteering to Traffic Dynamics in V2I Systems

The paper presents the simulation model and signal propagation environment to generate synthetic data and test trained intelligent algorithms for an intelligent beamsteering unit. This intelligent unit is for vehicle connection and tracking in a vehicle-to-infrastructure system simulation model. The system adapts the operation of the machine learning block to various traffic scenarios, improving the efficiency of the intelligent beamsteering system under all dynamic traffic conditions. We simulated eight training traffic models and five test models with different traffic scenarios to evaluate different approaches for ingenious block adaptation efficiency. We evaluated the efficiency of the decision-making system by conducting simulations for cases where only one decision tree was used, where we selected the active tree based only on the speed mode of connected vehicles, and where we used a Fuzzy Solver Block. When using Fuzzy Solve Block, the system shows higher stability at high traffic intensity values with an overall improvement in efficiency. The decision-making system showed a 1 dB reduction in the average signal-to-noise ratio deviation from the maximum value in the selected sector and a 6% probability of signal level reduction at the vehicle receiver input.