A task selection method for multi-faced static phased array radar resource management in dynamically changing environments using recomposable restricted finite state machines is presented. Restricted finite state machines allow the design of a composed finite state machine with resource limitations by restricting some of the inputs. Recomposable restricted finite state machines allow the state space of a finite state machine to change dynamically, which allows the modelling of a dynamically changing environment. Applying dynamic programming to restricted finite state machines yields optimal policies for a given cost function and applying breadth-first search or limited breadth-first search with fixed depth yields suboptimal solutions for the current state. The authors model a task selector for the radar in an overloaded battlefield situation using recomposable restricted finite state machines and obtain a radar resource allocation policy using dynamic programming when the environment changes dynamically and the resources are limited. The suboptimal solution for the current state is obtained using heuristic methods: breadth-first search, or limited breadth-first search in the task selector for large-scale problems. Furthermore, the authors consider distributed architectures for multi-radar systems with communication channels. The results show that their approach performs well from the standpoints of both computational time and performance.