The Question

Can the current Artificial Intelligence (AI) pathfinding methodology that is used for autonomous robots be improved to increase the accuracy of the robot reaching its destination? 


Current pathfinding algorithms work well for navigation in computer games, but not as well for navigating autonomous robots. Attempting to increase a robot’s accuracy, I developed an algorithm that balances pathfinding efficiency with accuracy using aspects of the robot’s surroundings (landmarks). Here is an overview of my algorithm:


Step 1: Configuration (done once):

Hospital map is uploaded

AI analyzes blockages

Makes routes between landmarks


Step 2: Nurse inputs destination and ideal route created via AI:

Nurse inputs destination

AI makes paths from start to each landmark and from each landmark to end

AI combines new paths with paths made in step 1

Combinations are made incrementally (one path at a time is added to a route)

AI chooses best combination by calculating a point score for each combination

AI generates executable instructions to destination based upon best route combination

Instructions are executed by the transporter


Point Score Determination:

Landmark usage is encouraged - points subtracted from score

Route length - points added to score

Turns add inaccuracy to the route - points added to score 

A multiplier is applied to distance values after each turn -- more turns = larger multipliers

Landmarks fix inaccuracies caused by turns and reset multiplier

Smallest scoring route is used


After developing my algorithm, I structured experimentation to test if my algorithm offers improved navigational accuracy over current AI pathfinding methods.

 

Project Summary
Find a question that interests you about something that you have observed, noticed or wondered about.


 

Judges' Tip
An excellent question will be interesting, creative, worded scientifically and relevant to the world today (250 words maximum).