Aimed at the problems that floating solution ambiguity is difficult precisely to deduce from realtime change of baseline parameters in dynamic to dynamic relative positioning， a method is proposed for dynamictodynamic integer ambiguity resolution based on Singular Value DecompositionRecursive Least Square (SVDPLS). In the method, the coefficient matrix of baseline vector is resolved by Singular Value Decomposition (SVD), double difference equation is transformed to eliminate baseline parameters， the floating solution and covariance matrix of double difference integer ambiguity are deduced in realtime by Recursive Least Square (RLS) method，and the ambiguity is searched and fixed by LAMBDA algorithm. The experiment result shows that the proposed algorithm can work out the single frequency integer ambiguity correctly in about 100 seconds, the baseline error within 1 centimeter, and is suitable for realtime solution to highprecision dynamictodynamic relative positioning.