Due to their high efficiency and robustness across various scenarios, multi-robot collaborative systems have emerged as a current research hotspot. However, a prerequisite for effective multi-robot collaboration is the calibration of relative positions among all nodes. Addressing the issue of relative position initialization relying excessively on external environments and motion stimuli, this paper focuses on the problem of two-dimensional relative position initialization based on ranging in environments where GNSS signals are denied and there is no external infrastructure assistance, with particular consideration given to constrained motion scenarios. By establishing a global coordinate system with the central node as the origin, the paper proposes a method to calibrate the absolute coordinates of all nodes in this coordinate system solely relying on two limited movements of the central node. Extensive simulations and experiments validate the effectiveness of the proposed relative localization initialization algorithm.