1. Crowd Dynamics AI Engine w/ PSO By: Keith Ainsworth Pd. 5

2. Crowd Dynamics AI in C++ ● Using a game for demonstration ● C++ with Object Oriented Programming ● SDL for graphical Display (might become OpenGL)

3. The AI: ● Crowd Dynamics ● Simulates the motion, movement, behavior of groups ● Conventional games do it individually ● Simple example: A group of people finding a target ● Particle Swarm Optimization

4. PSO ● A method to optimize the behavior of “swarms” ● Conventionally used in anthill simulations ● This only optimizes half of my AI engine ● Optimizes the movement calculations ● Does not optimize the communication aspect

5. Current Class Structure: Person PNetworkWorld weight_list

6. ● Looping, array-based, fixed-size container ● 2 lists, one for data, one for weighting in summation ● Summation functions, which return the average value of the list (taking weighting into consideration) ● Used by the PNetwork to pick a direction for a person to travel after receiving a bunch of tips from the other people

7. Pnetwork ● Instantiated with a list of people as the group ● Each time the step() method is called it  Calls the “communicator” method of each person  Adds them all into a weight_list for each person (differentiated by the weightings given)  Instructs the persons to travel in the direction of the summation of their weight_list

8. Simulation: ● Shows Regular Simulation

9. Subclassing Structure ● Program Built around base libraries which can be subclassed ● Classes include:  World (subclass: ngon_world)  Std_Person (subclass:theta_person, subclass of person)

10. ngon and ngon_world ● ngon is a class that essentially stores a polygon of n sides through the boundary points ● Can be concave, convex, even internally intersecting ● ngon_world is a subclass of world, which maintains the “global” variables to the simulation ● The subclass allows for obstructions in sight and movement of agents in the form of ngons

11. R_Limit.hpp ● The simulation wasn’t regular in it’s behavior in human/computer interaction ● Different computers = Different Difficulties ● This is a rate limiting library  Utilizes time.h system functions for maximum portability and accuracy  Two methods, queue() and hold(). Queue is called before the frame’s codes, hold is called at the end and holds if necessary to maintain a rate.  Slows a simulation to a rate but cannot speed it up to a rate it isn’t maintaining.

12. ngon_world