Accurate and fast estimation for states is the key to the multi-robot cooperative transportation. However

the majority of the existing multi-robot cooperative localization approaches have a common limitation in which they cannot satisfy the requirements to the positioning accuracy and computational complexity. According to the task characteristics of cooperative transportation

the rigid constrains of the range and azimuth information are first introduced into the cooperative localization. Moreover

the close coupling relationship between robots is fully utilized to establish the general and effective kinematics and measurement models with the rigid constrains. This facilitates the derivation of an efficient approach to the dual-robot cooperative localization based on Gauss-Hermite quadrature Kalman filter (QKF). Experimental results demonstrate that the proposed approach has much higher positioning accuracy than the QKF cooperative localization approach based on unconstrained models

and reduces the computational complexity largely. This paves the way for the real-time cooperative localization in practical applications.