In order to reduce the A* run time, we decide to split pathfinding into global planning and local collision avoidance components. We discuss how to model an agent-agent collision and reduce the problem into a linear optimization problem.
How does one scale up dozens of state mutation flows in an RTS game?
Mutating a shared state is especially hard if we lack a structured approach to change operations. We share the design doc DownFlux uses to answer this design question.
Because DownFlux is not built on top of any existing gaming engine, we need to design a way for client-server network connections to be resilient to network flakiness.
Outline the core mechanics necessary for rendering the server-calculated game state to the players.
Design a communications model between a small number of clients concurrently mutating a complex RTS game state.
Interesting papers I came across while researching RTS game architecture.