In order to remedy the position deviation of the joint angles and ensure the smooth motion trajectory of the end-effector at velocity level, a velocity-level tri-criteria optimization (VLTCO) scheme of redundant manipulator’s is proposed. The proposed VLTCO scheme combines the minimum velocity norm (MVN), the repetitive motion planning (RMP), and the infinity-norm velocity minimization (INVM) through the weighting factors, which guarantees the joint velocity to be near zero after finishing the redundant manipulators task. At the same time, the scheme considers the joint-angle and joint-velocity physical limits, which can keep the joint angle and joint velocity within their given range and prevent the occurrence of high joint velocity during the task duration. Finally, the VLTCO scheme is reformulated as one general quadratic program (QP) problem. The QP problem is then solved by a linear-variational-inequality-based primal–dual neural network solver (LVI-PDNN). The validity and advantages of the VLTCO scheme are substantiated by the simulation results and the physical experiment based on the JACO2 redundant manipulator.