Navigation Application of 3D Corner Features Measured with a 2D Laser Scanner

Zhen Zhu, Chris Venters, Maarten Uijt de Haag

Abstract:	This paper discusses the extraction of 3D corner features using 2D Laser scanners, and the potential applications of these corner features in an integrated navigation solution. Compared against a 3D flash Laser Detection and Ranging (LADAR), a 2D laser scanner has advantages in terms of cost, range and resolution. A 2D laser scanner can be used to sense the 3D surroundings when it is integrated with other sensors. It has found widespread applications in ground-based and airborne operations in natural, urban, and indoor environments. More particularly, the 2D LADAR is often integrated with an IMU in GNSS-denied environments, which is a similar process to the integration of other optical sensors (for instance, cameras) and RF sensors. In general, data from a given type of sensors can be fused into a navigation solution in various ways. For example, in a well-structured environment, features found in 2D laser data can help estimate the displacement and rotation between two scans. An "odometry" type of measurement will be generated to directly aid an IMU-based dead-reckoning system. It is in contrast to simultaneous localization and mapping (SLAM), which keeps track of a map of landmarks. SLAM and odometry algorithms may both be implemented in a filter framework, such as an EKF. In EKF-SLAM, the landmark information is stored, tracked and updated as a part of the filter's state space. A 3D feature extracted from LADAR scans could be a surface, a line, a point or in this paper, a corner. The feature would need to be identified and tracked across continued scans. Some point-features can better serve as landmarks in SLAM. When re-observed in multiple scans, these features are used to enforce a geometric constraint between the current states and history poses. In the meanwhile, the range from a LADAR to a planar surface or a line is obviously related to the instantaneous motion, and is a useful feature in odometry. Planar surfaces are abundant in a man-made structure. Two adjacent planes, such as two walls, can form a corner. Although corners in these structures are often close to having a right angle, the precise angle of a corner cannot be directly measured from a 2D laser scan. This phenomenon can be related to some of the LADAR calibration procedures. As discussed in this paper, there exists a deterministic geometric relationship involving the LADAR attitude, the 3D angle between both walls and the apparent angle observed in the 2D scan when the LADAR is tilted. As a result, a 3D corner could potentially provide direct and instantaneous observability to the attitude states in an EKF, which makes it an opportunistic feature. Furthermore, the location, orientation and angle of this 3D corner makes it a unique feature that can be easily recognized when revisited by the LADAR. Therefore, the corner could also be a great choice for SLAM landmarks. This paper will illustrate the abstraction of 3D corner features in 2D laser data, and explore the potential applications in odometry or SLAM solutions. The 3D corner features can be characterized in terms of the observability to the navigation states as an opportunistic feature or a landmark. The results will be presented with analytical studies, simulations and field test data.
Published in:	Proceedings of the 27th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2014) September 8 - 12, 2014 Tampa Convention Center Tampa, Florida
Pages:	2174 - 2182
Cite this article:	Zhu, Zhen, Venters, Chris, de Haag, Maarten Uijt, "Navigation Application of 3D Corner Features Measured with a 2D Laser Scanner," Proceedings of the 27th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2014), Tampa, Florida, September 2014, pp. 2174-2182.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In