This paper proposed a trajectory planning scheme combining the robot dynamics and constrained joint torque. With the objective of the robot′s steady movement, the trajectory is sampled equably and each segment is considered as uniformly variable motion. To guarantee tracking precision, the relation between the tracking control error and the output torque of the controller is analyzed from the point of engineering. Then an effective replanning scheme is presented, where fuzzy logic is used to weight the maximum joint torque according to the control error. The experiments on a robot show the feasibility of the scheme.