In the multi-UAV task planning problem, the curvature constraints of UAVs make the task execution solutions showing diversity, which brings challenges to the evaluation accuracy of task solutions and the allocation effectiveness. To address this problem, this paper designs a preceding-task trajectory matrix (PTTM) method as a novel decentralized task planning framework. The PTTM method adopts multiple optimal three-point trajectories as a fast approximation of the integral trajectory to estimate the overall reward of task sequences, to avoid extensive trajectory planning calculations, and a numerical solution is designed for the calculation of the elements in the PTTM model. Based on the consensus-based bundle algorithm, PTTM-CBBA algorithm is proposed, which includes the bidding procedure and market negotiation procedure. The simulation results show that compared with the original CBBA without PTTM, the algorithm proposed in this paper can obtain higher task rewards, that is, the multi-UAV system can achieve higher task execution performance.