Over the last few years, spatial partitioning systems have become critical for innovative AI such as node and cover-point analysis, team and squad management, and dynamic world awareness. On console platforms, however, the memory these systems use is a critical factor.
This lecture presents a new technique from the field of satellite imaging uniquely applicable to real-time spatial partitioning and path-finding. The technique centers around an adaptive method called "Vantage Point Modeling." A VP architecture is presented thats easy to understand and implement, radically reduces memory, is pointer-less and cache friendly and does not compromise performance.
The lecture also presents many tips and tricks that advance current AI through partitioning systems; topographic and world awareness, squad and team based AI, comrade awareness and advanced real-time cover analysis. Also covered; memory reduction tricks from the real world, splitting data for RAM/ROM based platforms and Bayesian particle sorting for node graphs.