A widely used technique for predicting traffic flows in individual roads of a road traffic network is the user-equilibrium (UE) traffic assignment (TA). This technique assigns trips from origins to destinations in a road traffic network so that all trips use the cheapest path. The cost of the path, which consists of roads (edges), is the sum of the roads costs. These costs increase with increasing flow in these roads. In this paper, we describe the parallelization of the B algorithm – a relatively new TA algorithm with a fast convergence to a solution. Since the nature of the algorithm and the nature of its fast convergence complicate the parallelization itself, we considered and implemented three parallel variants and tested them on real road traffic networks to investigate their convergence, usability, and speed. The parallelization is intended for a shared memory parallel computing environment. The description of the parallelization along with the performed tests is the main contribution of this paper.

一种广泛使用的用于预测道路交通网络各个道路中的交通流量的技术是用户平衡 (UE) 交通分配 (TA)。该技术在道路交通网络中分配从起点到终点的行程，以便所有行程都使用最便宜的路径。由道路（边）组成的路径成本是道路成本的总和。这些成本随着这些道路上流量的增加而增加。在本文中，我们描述了 B 算法的并行化——一种相对较新的 TA 算法，具有快速收敛到解的能力。由于算法的性质及其快速收敛的性质使并行化本身复杂化，我们考虑并实现了三个并行变体，并在实际道路交通网络上对其进行了测试，以研究它们的收敛性、可用性和速度。并行化旨在用于共享内存并行计算环境。并行化的描述以及执行的测试是本文的主要贡献。