A Game AI Approach to Autonomous Control of Virtual Characters
Bibliographic record
Abstract
There is a strong need to develop Artificial Intelligence (AI) for virtual characters which are: Autonomous – able to function effectively with little or no human input at runtime Reactive – aware of and responsive to the evolving situation and the actions of the trainees Nondeterministic – the viewer should never see exactly the same thing twice Culturally Authentic – act as a person of the portrayed culture would Believable – maintain immersion by acting in a believably human way This could greatly reduce the training costs, increase accessibility, and improve consistency. As one aspect of the Future Immersive Training Environment Joint Capabilities Technology Demonstration we created the “Angry Grandmother,” a mixed reality character portraying the elderly grandparent of an insurgent whose home is entered and searched by the trainees. She needed to be believable, culturally authentic, nondeterministic, and reactive within the limited scope of the scenario. In addition, she needed to be capable of autonomy, but also responsive to direction from the instructor/operator. The last 10 years have seen a dramatic improvement in the quality of the AI found in many video games; in our opinion, game AI technology has reached a level of maturity at which it is applicable to immersive training. Accordingly, we built an AI which combines Behavior Trees (BTs) and utility-based approaches. This approach is a descendant of that used in several extremely successful video games, including the Zoo Tycoon 2 franchise, Iron Man, and Red Dead Redemption. This paper will present the AI architecture which we used for the Angry Grandmother, compare and contrast it to relevant game AI approaches, and discuss its advantages particularly in terms of supporting rapid development of autonomous, reactive characters, but also in terms of enabling that crucial dichotomy between autonomy and operator control.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".