Game AI Action Model Using Generative Adversarial Imitation Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Game AI struggles to make decisions that align with human user expectations in complex and diverse video game environments, resulting in low intelligence levels and poor decision-making capabilities.
Innovation Solution
The implementation of a method that uses generative adversarial imitation learning to train game AI by observing and imitating real user actions, allowing the AI to learn a game action policy that matches user behavior, thereby enhancing its decision-making capabilities and personification effect.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If game AI is configured to imitate human game actions in complex video game environments, then the AI can interact with real users and provide game decision suggestions, but the AI struggles to make decisions that align with user expectations, resulting in low intelligence levels
Solution Approach 1:
The patent applies copying by training the game AI to imitate human user actions and decision-making patterns. The system collects data from real users playing the game and uses this data to train the AI to replicate human behavior, thereby improving the AI's decision-making capability and alignment with user expectations while maintaining adaptability to complex game environments
2Adaptability or versatility
If the game AI is exposed to diversified and increasingly complex game content, then the AI can interact with real users and provide suggestions, but the AI presents low intelligence level and poor decision-making capability
Solution Approach 1:
The patent applies preliminary action by pre-training the game AI on a comprehensive dataset of human user behaviors and decisions before the AI is deployed to handle complex game content. This pre-training process allows the AI to learn and internalize decision-making patterns from real users, enabling it to make intelligent decisions when encountering diverse and complex game scenarios without requiring real-time learning
3Ease of operation
If the game AI uses traditional decision-making approaches, then the AI can execute basic game actions, but the AI cannot meet user expectations for intelligent and personified behavior
Solution Approach 1:
The patent applies copying by training the game AI to imitate human user actions and decision-making patterns. The system collects data from real users playing the game and uses this data to train the AI to replicate human behavior, thereby improving the AI's decision-making capability and alignment with user expectations while maintaining adaptability to complex game environments
Data Source
AI summary
Embodiments of this application relate to the field of artificial intelligence technologies, and in particular, to an information processing method, an information processing apparatus, a computer readable storage medium, and an electronic device. The information processing method includes: determining, by a device, a subject in a game scenario, and acquiring an action model used for controlling the subject to execute a game action; performing, by the device, feature extraction on the game scenario to obtain model game state information related to the subject; performing, by the device, mapping processing on the model game state information by using the action model, to obtain model game action selection information corresponding to at least two candidate game actions; and selecting, by the device according to the model game action selection information, a model game action for the subject from the at least two candidate game actions.