Aiming at the problem of how hexapod robots make optimal human-machine command combinations in complex environments, it is proposed to divide the process of assigning control command permissions into two stages: offline and online. In the offline stage, the stability margin of the hexapod robot body, energy consumption, and collision contact time with obstacles during the motion process are used as constraints for command optimization, and an evaluation function is established to obtain the optimal control command combination before entering the terrain; In the online stage, the state transition probability in the finite state machine network is quantified based on the three constraints mentioned above, and the constraint threshold is manually set as the triggering condition to transition to the optimal human-machine decision instruction state. Finally, experiments are conducted using a semi physical simulation system, and the results show that the proposed decision method improves the stability margin by about 15%sim25% compared to other control methods; The energy consumption decreass by about 10%sim50%; The collision contact time is reduced by about 10%sim50%.