Aiming at the non-match and match situations of simulation robot fish, in order to solve the result is neither dependent on the choice of the initial line does not need outside intervention-specific circumstances, to achieve the fish fast, accurate adjustment, this paper proposed two kinds of ant colony algorithms action decision strategy. Ant colony algorithm based on branch bound method judges the key physical machine where the fish branch, self-determined speed and angular velocity in the moment and optimal combined speed of the fish with angular velocity of the fish; In each cycle, the ant colony algorithm based on dynamic programming make self adjustment according to the dynamics of the robot fish immediate feedback. Examples of the above two methods used by the 2D simulation platform validation results showed, robot fish can be adjusted based on the policy path, to achieve optimal combination of speed and direction, the shortest time and distance to find the target point. This shows that based on ant colony algorithm’s two kinds of action decision strategy has a strong ability to adapt effectively to meet the simulation of robot fish for action decision-making.