Abduction-Based speculative computation is a process which uses default hypothesis as tentative answer and continues computation when resources information cannot be obtained in time. In a computing process, if response is not consistent with belief, master Agent will revise its belief. To achieve goals, get the more accurate result of computation and reduce risks of decision under time constraints, the master Agent in speculative computation should make efforts to obtain more information via negotiation, so the negotiation is a key measure in reducing risks of decision. In this paper, basic theories of abduction and speculative computation are introduced. Based on these theories, a framework extended from the speculative computation with belief revision based on time constraints is presented, and the further negotiation framework based on time constraints and negotiation algorithm with belief revision are presented. The algorithm is embedded in speculative computation to make the master Agent revise its belief during negotiation. Finally, three experiments on goods transportation are given. The experimental results indicate that the algorithm can improve the accuracy of result of the speculative computation and reduce the risk of the result.