3DSE_Presentation

Download Report

Transcript 3DSE_Presentation

3D Spatial
Exploration
An artificial intelligence to pursue and locate a hidden target
Nebula 3
Has all the needed features
●
Real-time rendering
●
User input
●
Path-finding
●
GUI
AI Functionality
●
AI senses
–
●
Sight
AI behavior:
–
Probability based searching
–
Pursuing a visible target.
–
Detection of changes in the environment
Probabilistic seeking
●
●
●
●
●
Probabilistic Occupancy Map
Map/Grid of the world
Set probability to 1 where the target is seen.
Set probability to 0 where the target isn't
seen.
Probability will decay and spread, becoming
weaker as time goes on and it grows.
Probabilistic seeking
Sight
Do it as Thief?
●
Several view cones to simulate focus and
peripheral vision.
Synthetic vision?
●
Render the scene from the agents eyes.
Sight
How do we know where we've looked?
●
Probability Map?
–
●
Don't work in 3D (Hiding beneath objects).
Test view cones/frustum against an octree?
–
Mark cells as seen if X% is inside the view.
–
Otherwise, split into smaller cells
Needs more research and experimentation.
References
Isla D. & Blumberg B. - Object Persistence for
Synthetic Creatures
Leonard T. - Building an AI Sensory System:
Examining The Design of Thief: The Dark
Project