BDTM: A High-Performance Intelligent Transportation Information Transfer Model Based on Dynamic Adaptive Optimization Strategy
-
Abstract
In recent years, collaborative intelligent transportation systems (C-ITS) have been developing rapidly, and the efficient real-time processing of massive amounts of C-ITS information has become a new challenge. An information transmission model based on message queuing telemetry transport (MQTT) bridging to Kafka is a feasible solution; however, it faces significant performance bottlenecks due to fluctuations in upstream traffic. To address this issue, we propose BDTM, a high-performance information transmission model for intelligent transportation systems based on a dynamic adaptive optimization strategy. BDTM utilizes an MQTT-Kafka bridging architecture, leveraging the Kafka connector to enhance system fault tolerance and horizontal scalability. Building upon the MQTT-Kafka bridging architecture, BDTM integrates two core technologies. First, we propose a weighted Latin hypercube sampling (wLHS) algorithm improved based on the concept of linear regression. By quantifying the relative importance (weights) of key Kafka parameters on throughput performance, the wLHS algorithm dynamically optimizes the sampling coverage of the high-dimensional parameter space, generating an optimal configuration knowledge base while reducing modeling complexity. Second, we propose a dynamic adaptive optimization strategy based on the Autoregressive Integrated Moving Average (ARIMA) model for traffic forecasting. This strategy monitors and forecasts upstream Kafka traffic in real time and dynamically adjusts Kafka resource allocation based on the forecast results, achieving a precise match between computing resource supply and traffic demand. Experiments show that BDTM achieves a communication throughput increase of more than 20% compared with the original model, while reducing communication latency by more than 30%. Therefore, BDTM offers significant advantages in overall performance.
-
-