Abstract
Traditionally, search strategies are categorized into one of two types: explorative and exploitative. Explorative search strategies search by exploring an area without using gathered information to guide their search; these strategies function best when there is high clutter, which makes detections untrustworthy. Exploitative search strategies search by trusting available information and using it to guide their search; these strategies function best when there is low clutter, which makes detections, and thus, available information, more trustworthy. Infotaxis melds the two types of search strategies together by using explorative tactics in high clutter density and exploitative tactics in low clutter density to detect targets faster than established search strategies. It achieves this by maximizing information gain through maximizing entropy reduction. Infotaxis selects which grid cells it measures each iteration by calculating which cell will most reduce the amount of entropy or uncertainty. Previously, infotaxis has been used to search for a target with passive sensing. This thesis tests a version of infotaxis that instead uses active sensing to capitalize on available information and in turn, accelerate the search process. This thesis demonstrates that infotaxis is faster than traditional search strategies and can be implemented into a searching robot using ultrasonic sensors. Through constructing a MATLAB®® search strategy simulation, the speed of infotaxis is compared to three other search strategies: Maximum A Posteriori (MAP), cycling in order and random searching. Infotaxis is found to be faster than cycling in order and random searching under all tested conditions, and it is slightly faster than MAP when the probability of detection, 𝑃𝐷, decreases. Infotaxis is implemented into a real search for a single target by programming an iRobot Create® 2 to search a linear row of ten cells with the infotaxis method using an HC-SR04 ultrasonic sensor to make detections. With 𝑃𝐷=0.7 and a probability of false alarm, 𝑃𝐹𝐴=0.1, the robot finds the target with infotaxis 100% faster than MAP. By implementing infotaxis using a robot and ultrasonic sensor, this thesis demonstrates that infotaxis can be used as a search strategy in the real world, validating the simulations.