This paper studies an asymmetric confrontation problem, i.e., multiple agents with weak capability attack a powerful target. In particular, agents' autonomous decision-making is essential in a complex and situation-change scenario. A knowledge-based autonomous decision-making method is proposed for agents to improve their decision-making ability in the confrontation. The proposed method is based on the tactical pursuit point (TPP) mechanism and each agent's current situation. Firstly, each agent conducts a situation assessment based on the available information, including situation type assessment and situation superiority assessment. Then, the TPP mechanism is utilized to generate a continuous tactical pursuit space (set of all possible tactical pursuit points). To reduce the complexity of decision-making, the tactical pursuit space is defined by a tactical pursuit basis and a decision-making vector. In addition, the decision-making vector is calculated by a designed function based on the situation assessment information. Therefore, a tactical pursuit point can be determined by the decision-making vector in the tactical pursuit space. Finally, under the guidance of the tactical pursuit point, each agent performs the attack maneuver. Simulation results show that the agents can make autonomous decisions during the confrontation. Meanwhile, multiple agents cooperate to attack a target, and the number of casualty agents is reduced.