Efficient scene simulation for robust monte carlo localization using an RGB-D camera

Abstract

This paper presents Kinect Monte Carlo Localization (KMCL), a new method for localization in three dimensional indoor environments using RGB-D cameras, such as the Microsoft Kinect. The approach makes use of a low fidelity a priori 3-D model of the area of operation composed of large planar segments, such as walls and ceilings, which are assumed to remain static. Using this map as input, the KMCL algorithm employs feature-based visual odometry as the particle propagation mechanism and utilizes the 3-D map and the underlying sensor image formation model to efficiently simulate RGB-D camera views at the location of particle poses, using a graphical processing unit (GPU). The generated 3D views of the scene are then used to evaluate the likelihood of the particle poses. This GPU implementation provides a factor of ten speedup over a pure distance-based method, yet provides comparable accuracy. Experimental results are presented for five different configurations, including: (1) a robotic wheelchair, (2) a sensor mounted on a person, (3) an Ascending Technologies quadrotor, (4) a Willow Garage PR2, and (5) an RWI B21 wheeled mobile robot platform. The results demonstrate that the system can perform robust localization with 3D information for motions as fast as 1.5 meters per second. The approach is designed to be applicable not just for robotics but other applications such as wearable computing.

Incremental and batch planar simplification of dense point cloud maps

“…The polygon-based and point-based methods offer a trade-off depending on the desired number of triangles or the intended use of the final triangulation. With robot navigation in mind, the low polygon-count models achieved with our system are suitable for use in a primitives-based localization system, such as the KMCL system of Fallon et al [4].…”

“…In addition to this, some features of real-world maps, such as walls and floors, end up being over-represented by thousands of points when they could be more efficiently and intelligently represented with geometric primitives. In particular the use of geometric primitives to represent a large 3D map to localise against has been demonstrated as a feasible means of real-time robot localisation [4]. In this paper, we examine the problem of planar surface simplification in large-scale point clouds with a focus on quality and computational efficiency in both online incremental and offline batch settings.…”

Improvement of 3d Monte Carlo Localization Using a Depth Camera and Terrestrial Laser Scanner

“…However, previous work of MCL has been mostly limited to 2D motion estimation in a planar map using 2D laser scanners (Dellaert et al, 1999;Thrun et al, 2001). Recently, a few 3D MCL approaches have been proposed where rough 3D models and consumer-level depth cameras are used as the environment maps and outer sensors (Fallon et al, 2012;Hornung et al, 2014;Jeong et al, 2013). However, their localization accuracy and efficiency still remain at an unsatisfactory level (a few hundreds millimetre error at up to a few FPS) (Fallon et al, 2012;Hornung et al, 2014), or the accuracy is not fully verified using the precise ground truth (Jeong et al, 2013).…”

“…Recently, a few 3D MCL approaches have been proposed where rough 3D models and consumer-level depth cameras are used as the environment maps and outer sensors (Fallon et al, 2012;Hornung et al, 2014;Jeong et al, 2013). However, their localization accuracy and efficiency still remain at an unsatisfactory level (a few hundreds millimetre error at up to a few FPS) (Fallon et al, 2012;Hornung et al, 2014), or the accuracy is not fully verified using the precise ground truth (Jeong et al, 2013). Therefore, the purpose of this study is to improve an accuracy and efficiency of 6DOF motion estimation in 3D Monte Carlo Localization (MCL) for indoor localization.…”

Information-Driven Adaptive Structured-Light Scanners

“…Range sensors have revolutionized computer vision in recent years, with commodity RGB-D scanners allowing us to easily tackle challenging problems such as articulated pose estimation [27], Simultaneaous Localization and Mapping (SLAM) [16,31,6], and object recognition [15,21]. The use of 3D sensors often relies on a simplified model of the resulting depth images that is loosely coupled to the photometric principles behind the design of the scanner.…”