Despite the success of the Roomba robotic vacuum, similar robotics platforms have been slow to follow. Roomba’s success lay in its promise of accomplishing a simple repetitive task (vacuum cleaning) autonomously and at a competitive price.We present a low-cost robot platform which, like the Roomba, is designed to perform a specific task (picking up balls during a tennis match) autonomously and reliably. A tennis “ball boy” robot must be more aware of the environment than a Roomba. Rather than wandering about blindly, it needs to be able to find its own location and navigate to different points of interest as the situation warrants. Specifically, it needs to stay off the court during a match until it recognizes a dead ball, fetch the ball, and re-park itself at the edge of the court.This platform is self-contained with on-board sensing and computation, uses only cost-effective off-the-shelf components, and was designed to perform robustly and repeatably in semistructured real-world environments. This paper presents a particle-filter implementation that allows the robot to localize itself and navigate on a tennis court. Along with navigation and tennis ball detection, this will enable it to pick up tennis balls on a real tennis court.In this paper, we present our first steps toward this goal and analyze some of the challenges involved. First we constructed a mobile robot platform with a ball pickup attachment, and implemented a ROS software stack for base control and odometry. We then extended the capabilities of the software stack with a motion planner, a court localization module, and a stationary ball detector. To read this external content in full, download the paper from the author archives at the University of California, Berkeley.