Localization, Mapmaking, and Distributed Manipulation with Flexible, Robust Mobile Robots

Abstract

Current mobile robots (mobots) have extremely limited usefulness in real-world applications, due primarily to low capability and low reliability. Current mobots cannot navigate in environments which have not been extensively engineered to be mobot-friendly; they are also incapable of performing useful duties at their destination. They often have poorly designed hardware and primitive software-development environments, making the development of new mobile robot protocols slow and difficult. In this thesis, we describe three ways to make mobile robots more usable. We present algorithms for mobile robot self-localization, a design paradigm for reliable mobile robots, and protocols for cooperative large-scale manip- ulation by mobile robots. Localization is the process of determining the robot's location within its environment. More precisely, it is a procedure which takes as input a geometric map, a current estimate of the robot's pose, and sensor readings, and produces as output an improved estimate of the robot's current pose (position and orientation). We describe an algorithm which performs mobile robot localization using a geometric model of the world and a point-and-shoot rang ing device. We also describe a rasterized version of this algorithm which we've implemented on a real mobile robot equipped with a laser rangefinder we designed. We next focus on the mobile robots themselves. We have designed and built several mobile robots at Cornell, that feature robustness, flexibility, and ease of programming. Using a modular design approach, we've attained an unusual state for a university robotics lab: our mobots are almost always functional. In their basic configuration, the robots can perform many tasks; also, it is easy to add sensors and actuators to our mobile robots, allowing their use in many different applications (e.g. manipulation). We are interested in protocols for manipulation of large objects (e.g., boxes, wheeled carts) by cooperating mobile robots, particularly protocols that are asynchronous, on-line, and use no communication between mobots. We've developed a Pusher-Steerer model for cooperative manipulation, which enables two mobots to manipulate objects through complex paths. We describe and anal yze the model and describe its performance in several real experiments.