As the development of single robot SLAM technology advances, research on multiple robots is also actively progressing. In this paper, we propose an algorithm that estimates the relative pose between multiple robots. This study mainly focuses on estimating the relative initial orientation. In order to estimate the relative orientation through scan matching, we propose a method of using the polar histogram that converts the scan data into the angle and range domain as an initialization for scan registration algorithms. Based on the greedy algorithm, the polar histogram method finds the relative orientation that yields the smallest Euclidean distance between polar histograms of two scans. The experimental results show that the initial value from the proposed algorithm yields better performance than that without an initialization. Compared to other registration algorithms, the Normal Distributions
Transform (NDT) benefits more from the proposed initialization method.