Sophisticated sensors allow robots to see and hear the world at a level far beyond humans, but when it comes to interpreting the data they’re still a few notches below Daleks. Scientists at Duke University and the University of New Mexico have used the game “Marco Polo” as the inspiration for the creation of an algorithm that allows robots to identify and intercept moving targets.
Like in Marco Polo - a tag game often played in a swimming pool where the players have to seek out others using sound not sight - the robots will study changes in the sensory output of their targets, and use the information to predict where their targets are likely to go. Robots will then calculate the quickest possible route to the space where the target is most likely to be.
So far, the algorithm has allowed robots to apprehend targets moving in a straight line at a constant speed, but the researchers hope to develop it to the point where robots can compensate for a target’s evasive action.
Before developing the Marco Polo algorithm, the researchers created a “cell decomposition” algorithm that allowed robots to divide space into distinct parts, and navigate through areas without hitting obstacles.
The new algorithm links robots to a variety of stationary sensors, ensuring that all “cells” in an area are under constant surveillance. The large amount of sensory data will also allow them to make more precise judgments about a target’s trajectory.
While the constant talk of intercepting targets sounds quite sinister, robots with the Marco Polo algorithm would have a large variety of applications outside security. In addition to catching intruders, the system could be used for tracking endangered species, understanding the behavior of herd animals, and monitoring and addressing radiation leaks, chemical spills, and mines.
"The idea is that multiple sensors are deployed in the space to cooperatively detect moving targets within that space," said Rafael Fierro, associate professor of electrical engineering at the University of New Mexico. "As the sensor makes more detections, it is better able to predict the likely path of the intruder. The ultimate path taken by the robot sensor is one that maximizes the probability of detection and minimizes the distance needed to capture the target."
The results from the latest experiments are published in the latest issue of Journal on Control and Optimization, a publication of the Society for Industrial and Applied Mathematics.