The deep spatio-temporal data model network structure is mainly used in research fields that contain a large amount of spatio-temporal data. Currently, the typical application example is traffic management. However, a large amount of spatio-temporal data can also be generated in military confrontation fields.Traditional military confrontation simulation platforms mainly study the operating mechanism inside weapon systems, set certain command strategies and communication rules, simplify time and space factors, focus on improving the intelligence of combat groups, and lack attention to regular patterns and intelligence in time and space dimensions of individuals. An appropriate deep spatio-temporal data model network structure can be used to fully utilize the spatial and temporal factors, discover more intelligent expression patterns of complex confrontational individuals in military confrontation, improve the autonomous ability of combat individuals, and better serve future intelligent warfare.This paper establishes multi-spatio-temporal agent confrontation simulation platform based on the deep spatiotemporal data model and military simulation applications. First, the requirements of the military confrontation simulation platform are analyzed. Then, specific task scenarios are given to simulate the behavior of all independent spatio-temporal agents in the entire military confrontation. This provides a reference for commanders to make effective decisions.