Dog Driven Robot: Towards Quantifying Problem-Solving Abilities in Dogs

Abstract

Training a working dog for a role such as bomb detection or search and rescue can take years, incurring a large cost. As detection work requires visual and odor-based problem-solving abilities, we have started to investigate how to quantify these abilities to assess a dog's potential. Towards this goal, we studied if and how a dog can remotely drive a robot through a simple maze. In our feasibility study, the dogs interact with an affordance mounted on a raised platform which triggers robot movement while also providing visual feedback from the robot. This study evaluated three affordances (tug, button press, and proximity) on two participants to drive the robot through a straight course. By identifying the best robot-canine interaction method, we show the potential for canines extending visual problem-solving abilities to an external device.