Partial Replanning for Natural Character Actions in Video Games
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video game technologies using hierarchical task networks for character control often result in unnatural actions when task execution fails, as they require replanning from the beginning, leading to unexpected and unrelated actions.
Innovation Solution
A system that extracts tasks from a domain with compound actions, determines execution orders, and controls character actions based on a generated plan, with a partial planning function that excludes failed methods and replans using compound tasks as starting points to maintain natural action sequences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If replanning is executed from the beginning when task execution fails, then the planning process can be restarted, but unnatural actions appear because unrelated actions are suddenly executed
Solution Approach 1:
The patent segments the replanning process by identifying the failure point in the task network and restarting planning only from that specific node rather than from the beginning. This is achieved by detecting which task or method failed and using it as the new starting point for plan generation, thereby maintaining continuity of preceding actions while correcting the failure locally.
Solution Approach 2:
The patent performs preliminary analysis of the task execution state to identify where the failure occurred before initiating replanning. By examining the execution history and determining the specific failure point in advance, the system prepares an targeted replanning approach that avoids unnecessary re-planning of successful actions, thus maintaining natural character behavior.
2Adaptability or versatility
If a hierarchical type task network is used for character control, then optimal tasks can be selected according to World State, but the process becomes complex when replanning is needed
Solution Approach 1:
The patent divides the complex hierarchical task network into manageable segments by identifying and isolating the failure point. Instead of reprocessing the entire task network, only the failed task and its dependent tasks are re-planned, significantly reducing computational complexity while maintaining adaptability to the current World State.
Solution Approach 2:
The patent applies partial replanning by performing planning operations only on the necessary portion of the task network affected by the failure. This partial action approach avoids the excessive computation of full replanning while ensuring that the character's actions remain adaptive and natural.
Data Source
AI summary
A non-transitory computer-readable medium including a video game processing program is provided. The video game processing program causes a server to: extract tasks from a domain including at least one compound task, methods being set to each of the at least one compound task; determine tasks to be executed from the tasks in addition to an execution order thereof, and generate the plan; control an action of the character on a basis of the plan; and exclude, when the task is a compound task to which methods are set in a situation that the character fails the task in a middle of the plan, a failed method from a selection target, and execute partial replanning to obtain a partial modification plan. The action of the character is controlled on a basis of the partial modification plan in a case where the partial modification plan is obtained.