The paper presents two different methods for mobile robot tracking and following of a fast-moving person in outdoor unstructured and possibly dynamic environment. The robot is equipped with laser range-finder and omnidirectional camera. The first method is based on visual tracking only and while it works well at slow speeds and controlled conditions, its performance quickly degrades as conditions become more difficult. The second method which uses the laser and the camera in conjunction for tracking performs well in dynamic and cluttered outdoor environments as long as the target occlusions and losses are temporary. Experimental results and analysis are presented for the second approach

Detection of obstacles for autonomous vehicles is more difficult when the terrain is not locally planar and remains an open problem. We have developed an approach suited for obstacle detection in those cases where the terrain has significant curvature but is smooth enough that the obstacles are discrete. Our system consists of a low-cost scanning laser rangefinder and a novel algorithm that can reliably detect obstacles as small as 15 cm in curving terrain. This paper presents an analysis of the effectiveness of our system and a summary of experimental results from an outdoor mobile robot.

This paper presents a comprehensive navigation system capable of extended coverage operations in semi-structured environments. By semi-structured, we mean pre-mapped terrain that although locally smooth has potentially large changes in elevation and that is generally free of obstacles. The system has three key capabilities: it is able to track specified paths with high accuracy, detect small obstacles reliably, and plan coverage patterns to completely cover a specified area. These technologies have been combined and implemented on a mobile robot, which has accumulated over 90km of autonomous operation to date. Here we report on the components, the architecture, and experimental results.

Obstacle detection is a key component of autonomous systems. In particular, when dealing with large robots in unstructured environments, robust obstacle detection is vital. In this paper, we describe an obstacle detection methodology which combines two complementary methods: adaptive color segmentation, and stereo-based color homography. This algorithm is particularly suited for environments in which the terrain is relatively flat and of roughly the same color. We will show results in applying this method to an autonomous outdoor robot.

We describe the Cye personal robot, concentrating on its path planning and navigation. We start out with a brief description of the general problem, followed by more detail on the Cye robot. After that, we concentrate on the path planner, which uses a global potential field approach, combined with a novel optimization criterion. This ensures that the robot maintains adequate distance from obstacles, while finding an optimal path which incorporates partial terrain knowledge. This system has been in use for over five months, with approximately 200 installations. As we empirically show, it is robust and can handle dynamic obstacles and imperfect maps.

Many current avenues of vision research involve fully analyzing an image with expensive, high powered computers. This approach has major implications in terms of cost, size, and power consumption. Other methods have involved sub-sampling an image to reduce cost and complexity. This has the disadvantage of information loss. We present a low cost, low powered, reduced complexity vision system capable of intelligently sampling an image to reduce this information loss. The design philosophy and methodology is discussed, along with sample applications. Primarily we demonstrate how the reduced complexity vision system will be used to aid in navigation of an autonomous flying vehicle. This is quantified by showing how having multiple sampling schemes result in increased robustness and accuracy of our helicopter line tracking algorithm.