This paper presents and empirically compares two solutions to the problem of vision and self-localization on a mobile robot. In the commonly used particle filtering approach, the robot identifies regions in each image that correspond to landmarks in the environment. These landmarks are then used to update a probability distribution over the robot's possible poses. In the expectation-based approach, an expected view of the world is first constructed based on a prior camera pose estimate. This view is compared to the actual camera image to determine a corrected pose. This paper compares the accuracies of the two approaches on a test-bed domain, finding that the expectation-based approach yields a significantly higher overall localization accuracy than a state-of-the-art implementation of the particle filtering approach. This paper's contributions are an exposition of two competing approaches to vision and localization on a mobile robot, an empirical comparison of the two methods, and a discussion of the relative advantages of each method.
James Bruce Tucker Balch Manuela Veloso School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 Abstract Vision systems employing region segmentation by color are crucial in applications such as object tracking, automated manufacturing and mobile robotics. Traditionally, systems employing realtime color-based segmentation are either implemented in hardware, or as very specific software systems that take advantage of domain knowledge to attain the necessary efficiency. However, we have found that with careful attention to algorithm efficiency fast color image segmentation can be accomplished using commodity image capture and CPU hardware. This paper describes a system capable of tracking several hundred regions of up to 32 colors at 30 Hertz on general purpose commodity hardware. The software system is composed of three main parts; a color threshold classifier, a region merger to calculate connected components, and a separation and sorting system to gather various region features and sort them by size. The algorithms and representations will be described, as well as descriptions of three applications in which it has been used.
This paper presents and empirically compares solutions to the problem of vision and self-localization on a legged robot. Specifically, given a series of visual images produced by a camera on-board the robot, how can the robot effectively use those images to determine its location over time? Legged robots, while generally more robust than wheeled robots to locomotion in various terrains (Wettergreen & Thorpe 1996), pose an additional challenge for vision, as the jagged motion caused by walking leads to unusually sharp motion in the camera image. This paper considers two main approaches to this vision and localization problem, which we refer to as the object detection approach and the expectationbased approach. In both cases, we assume that the robot has complete, a priori knowledge of the three-dimensional layout of its environment.