Multi-agent planning system for controlling non-player agents in a game

An apparatus, comprising:

• at least one hardware processor;
• and
• computer memory holding computer program instructions executed by the hardware processor, the computer program instructions comprising program code configured to:
• receive a data set;
• transform the data set into re-composable combinations of tasks structured as a multi-modal hierarchical network of tasks, wherein an individual path in the hierarchical network of tasks represents a plan having a starting point and that represents a particular implementation of a task that has one or more ways of being implemented;
• and
• during a session of a game being controlled by a game engine associated with the game platform and involving a player, real-time orchestrate behaviors of one or more non-player agents by receiving a set of observations occurring in the game and, in response, implement a set of one or more plans in the hierarchical network of tasks, wherein orchestration occurs remotely from the game engine;
• wherein orchestrating behavior of an given agent includes receiving an observation message representing a game event, searching the multi-modal hierarchical network of tasks to identify a particular plan, based on one of:
• the identified particular plan, and an adaptation of the identified particular plan, generating a set of one or more cues, and providing the set of one or more cues to control the game engine in real-time.