Upload
jianfeng-chen
View
230
Download
0
Embed Size (px)
Citation preview
Robot Arena Game—— CSC584 Project Presentation
Yao LuRui Zhi
Jianfeng Chen
Inspiration — Robot Sumo
Blueprint
I don’t wanna die...T_T
Someone help...
I wish I could be smarter..
Architecture Overview● World Representation
○ Arena Map○ Robot Character○ Physics Simulation
● Individual AI○ Behaviors○ Decision Tree○ Evaluation
● Group AI○ Defense-based Strategy
Arena Map (Overview)
Arena Map (Things Behind)
Shelter Points
Arena Map (Things Behind)
Dirichlet Domains
Arena Map (Things Behind)
Walls and Borders
Robot Character
Scout
Tank
Soldier
Hero
Commander
Attributes: mass, fuel, status, vision, auditory
Physics Simulation (Overview)● Ground Friction● Rigid Body Collision● Wall Collision● Player Control● Ray-casting/Pathfinding
Video Time!
1. physics 2. ray-casting
Architecture Overview● World Representation
○ Arena Map○ Robot Character○ Physics Simulation
● Individual AI○ Behaviors○ Decision Tree○ Evaluation
● Group AI○ Defense-based Strategy
Individual AI (Behaviors)
Individual AI (Behaviors)
Video Time!
(behaviors)
Guard
● Attack● Evade● Wander● Hide● Buff● Ultimate● Trace● Tackle● Guard
Individual AI (Decision Tree)
● Enemy visible?● Enemy audible?● Enemy distance?● Enemy stronger?● Enemy VIP?
● Enemy Direction?● Ally Detectable?● Low fuel?● Is defensive?● Buff distance?
Individual AI (Evaluation)
Individual AI (Evaluation)
Architecture Overview● World Representation
○ Arena Map○ Robot Character○ Physics Simulation
● Individual AI○ Behaviors○ Decision Tree○ Evaluation
● Group AI○ Defense-based Strategy○ Evaluation
Group AI (Defense-based Strategy)
When one robot is in danger, its teammates should consider providing support.
ASI (Ally Safety Index) = ∑ (allyMass / reachTime)HSI (Hide Safety Index) = selfMass / reachTime
ETI (Enemy Threat Index) =∑ (enemyMass / reachTime) BTI (Border Threat Index) = 1 / reachTime
TSI (Total Safety Index) = ASI + HSI - ETI - BTI
Group AI
When one robot is in danger, its teammates should consider providing support.
Path Risk (teammate) = ∑ (enemyMass / reachTime)Gain (teammate) = TSI (seek) - TSI (current)
Group AI (Defense-based Strategy)
Group AI (Defense-based Strategy)● Defense-based Strategy
When one robot is in danger, its teammates should consider providing support.
for all Gain(robot) > 0 dowhile(ally still in danger) doGainMAX(robot) -> go guard the one in danger
endend
Group AI (Evaluation)
Show Time!
Q & A