Rapid Prototyping with AI Agents

Using Unity to animate autonomous agents via GOAP (Goal Oriented Action Planning) allows for a rapid-testing iterative study of circulation patterns in a workout space. A study of the circulation patterns of AI agents in a standard rectilinear gym reveals the possibility for a spatial separation of user groups. A series of tests yields a new floor plan wherein exercisers, traveling along chords of an arc, maintain distance from each other and from employees that travel along point-to-point radial paths on the opposite side of the arc.

Iteration 1

Exercising AI agents (pink trails) cross over each others’ exercise machines too frequently.

Iteration 2

A revised layout yields fewer exerciser-on-exerciser crossovers and fewer (contagious) interactions, but exercisers cross over into the employees half of the space.

Iteration 3

Exercisers and employees stay on their respective halves of the room, but exercisers continue to cross through each others’ space.

Iteration 4

Iteration 4 takes advantage of point-to-point movement by exercisers along chords of an arc, and radial point-to-point movement by employees.

Iteration 5

Using Iteration 4 as a template, the floorplate is optimized. When agents enter the space they are effectively sorted by user-type (based on programmatic needs) to one side of the room or the other, and interactions between members of distinct user groups are limited as much as programmatic constraints allow.