My experiment tests the accuracy of the resulting paths of two different pathfinding algorithms by measuring the accumulated variance of the robot's ending position from its target destination, in feet. 

I used the following variables and controls:

Independent Variables: Path plotted by the AI algorithms used 

Control group: Commonly used game AI algorithm (A-star)

Experimental group: Accuracy/efficiency balancing algorithm

Dependent Variable: Accumulated variance of the robot's ending position from its target destination measured in feet.

Controlled Variables: 

1. The same simulated hospital environment was used to test both algorithms.

2. Simulated robots were used to test both algorithms and eliminate environmental influences such as: 

a. Robot's power level

b. Friction between the robot and the floor

c. Robot's mechanical performance

3. Both robots have the same starting location and target destination.

4. Through my observations during my eight years as part of a competitive robotics team, I found that the accuracy of autonomous robots varies the most when the robot makes turns. To account for this factor, I tested a robot and calculated (see data section for details) a turn variance to apply to turns made by the robot in my simulations. This turn variance is applied equally to both simulations.

5. No variance was applied to the initial turn performed prior to the first segment of the path, eliminating possible path favoritism due to robot's initial direction. 

Procedure: Maintaining controlled variables, I applied each algorithm to identical VA Hospital scenarios, and obtained the resulting plotted paths (in data section). The simulated robot starts at the docking station and moves, while avoiding the lab and nurses' station, to deliver clean linens to the patient's bedside located at the yellow starred area. The colored lines are decorative floor tiles located in the hospital. 

Materials List: 

 One Computer to run algorithms

 Game AI algorithm (A-star)

 Accuracy/efficiency algorithm

 Paper, pencil, and hospital map to plot paths

 Ruler to measure accumulated variance of the robot's ending position from its target destination

Algorithm testing procedure:

Step 1: Input testing parameters, including:

Robot's starting location

Target destination

Locations of obstacles (things to avoid) and landmarks (walls, floor tiles, etc. that can be used for guidance).

Robot's size

Robot's starting direction

Step 2: Run both algorithms separately, using testing parameters.

Step 3: Plot paths generated by algorithms on separate maps.

Step 4: For both algorithms, measure the accumulated variance of the robot's ending position from its target destination in feet (dependent variable). 

Step 5: Repeat entire testing procedure four additional times to ensure accuracy.


I worked first with my control group, the game AI algorithm (A*), which doesn't use landmarks and views the hospital this way:

Each square represents approximately one square foot. I then tested my experimental group - my algorithm, which uses landmarks and views the hospital this way:

Only the walls and darkest floortiles were provided to the algorithm, which chose to use three floortiles and one wall. The results for both experiments are in the data section.


Design and execute an experiment that tests your hypothesis. Include descriptions of the materials, equipment, and methods/techniques you used. Explain the variables and how they will be controlled, manipulated and measured. Also detail any key steps to avoid errors, risks and safety


Judges' Tip
Excellent students will demonstrate that they have used good experimental techniques and describe their experiment clearly and in detail (500 words maximum).