Tracking Naturally Occurring Indoor Features in 2-D and 3-D with Lidar Range/ Amplitude Data

Abstract

Sensor-data processing for the interpretation of a mobile robot's indoor environment, and the manipulation of this data for reliable localization, are still some of the most important issues in robotics. This article presents algorithms that determine the true position of a mobile robot, based on real 2-D and 3-D optical range and inten sity data. We start with the physics of the particular type of sensor used, so that the extraction of reliable and repeatable information (namely, edge coordinates) can be determined, taking into account the noise associated with each range sample and the possibility of optical multiple-path effects. Again, applying the physical model of the sensor, the estimated positions of the mobile robot and the uncertainty in these positions are determined. We demonstrate real experiments using 2-D and 3-D scan data taken in indoor environ ments. To update the robot's position reliably, we address the prob lem of matching the information recorded in a scan to, first, an a priori map, and second, to information recorded in previous scans, eliminating the need for an a priori map.

CC-RANSAC: Fitting planes in the presence of multiple surfaces in range data

“…A simple categorization of range analysis algorithms may be drawn based on whether local or global descriptors are employed. Local descriptors include local surface normals [7,8], ridges [9], and discontinuities [10,11,12,13]. Contiguous point sets with similar local descriptors may be clustered in space in order to identify extended regions.…”

Robust curb and ramp detection for safe parking using the Canesta TOF camera

“…Local descriptors include local surface normals [18,24], ridges [10], and discontinuities [32,34,4,5]. Contiguous point sets with similar local descriptors may be clustered in space in order to identify extended regions.…”

Dynamic Environment Exploration Using a Virtual White Cane

“…Hence, it is conceivable that by modeling surfaces as piecewise planar, and by tracking the range using suitable models, discontinuities between surfaces can be detected and characterized. A similar problem have been studied for robotics navigation, with surfaces being tracked by a rotating lidar [2,1]. However, whereas a lidar takes calibrated point measurements (i.e., the angular period between two consecutive measurements is known), the velocity at which the device is being pivoted, as well as the pivoting direction, are unknown to the processing module in our case.…”

“…We thus propose to substitute the terms in the first line of (9) with an approximation of x n−1 . More precisely, rather than branching on the condition x n−1 (1) > x n−1 (2), we will use either the condition x n−1 (1) >x n−1 (2) , wherex n−1 is the estimate of x n−1 , or the condition z n−1 > z n−2 on the observations.…”

“…Thus, the user will be able to explore the environment with a very narrow laser beam receiving information about features and their distances, and integrating it into a mental image of the scene. Note that laser technology allows for high spatial and temporal resolution that are far superior to those of existing ETA's based on ultrasound technology 2 . In order to validate our ideas, we have built a simple proof-of-concept prototype [18], shown in Figure 1 and briefly described in Section 2.…”