Search

In this paper, we present a multirobot exploration algorithm that aims at reducing the exploration time and to minimize the overall traverse distance of the robots by coordinating the movement of the robots performing the exploration. Modeling the environment as a tree, we consider a coordination model that restricts the number of robots allowed to traverse an edge and to enter a vertex during each step. This coordination is achieved in a decentralized manner by the robots using a set of active landmarks that are dropped by them at explored vertices. We mathematically analyze the algorithm on trees, obtaining its main properties and specifying its bounds on the exploration time. We also define three metrics of performance for multirobot algorithms. We simulate and compare the performance of this new algorithm with those of our multirobot depth first search (MR-DFS) approach presented in our recent paper and classic single-robot DFS.

A great variety of biological groups form a self-organized swarming motion at some point during their life spans, which has two prominent collective features: common velocity and constant spacings among members. In this paper, we present a general individual-based motion framework to explain such collective motion of swarms in homogeneous environments. The motion framework utilizes the concept of social interactions that has been widely accepted throughout the literature. We assume that during the motion of the swarm, each member senses and interacts with its neighbors via virtual Attraction/Alignment/Repulsion (A/A/R) forces, while perceiving and following the gradient force of the environment. During the swarm's motion, the neighborhood and the interaction relations among members may dynamically change. To explicitly consider the effect of such dynamic change on the emergence of swarm's collective behavior, we use an algebraic graph to model the topology of the interaction and the neighborhood relations among the members. By using mathematical tools of nonsmooth analysis theory and Lyapunov stability theory, we analytically prove that if the A/A/R forces have limited ranges, and the attraction/repulsion forces are balanced at a certain range, the proposed framework leads to a parallel type of collective motion of the swarm. We mathematically show that the velocities of all swarm members asymptotically converge to a common value and the spacings among neighbors remain unchanging. In addition to the mathematical analysis, a few sets of simulation results are included to demonstrate the presented framework. The contributions of this paper are twofold: First, unlike most works in the literature that mainly use computer simulations to study the swarming phenomena, this paper provides an analytical methodology to investigate how the collective group behavior is self-organized by individual motions. Second, the presented motion framework works over a general range of A/A/R interactions. In other words, we analytically prove that the commonly used A/A/R model can lead to a collective motion of the swarm. In addition, we show that the alternative model in the literature that uses only attraction/repulsion (A/R) interactions is in fact a special case of the A/A/R model., (Copyright (c) 2010 Elsevier Ltd. All rights reserved.)

In this paper, we present a system for estimating the trajectory of a moving RGB-D camera with applications to building maps of large indoor environments. Unlike the current most researches, we propose a 'feature model' based RGB-D visual odometry system for a computationally-constrained mobile platform, where the 'feature model' is persistent and dynamically updated from new observations using a Kalman filter. In this paper, we firstly propose a mixture of Gaussians model for the depth random noise estimation, which is used to describe the spatial uncertainty of the feature point cloud. Besides, we also introduce a general depth calibration method to remove systematic errors in the depth readings of the RGB-D camera. We provide comprehensive theoretical and experimental analysis to demonstrate that our model based iterative-closest-point (ICP) algorithm can achieve much higher localization accuracy compared to the conventional ICP. The visual odometry runs at frequencies of 30 Hz or higher, on VGA images, in a single thread on a desktop CPU with no GPU acceleration required. Finally, we examine the problem of place recognition from RGB-D images, in order to form a pose-graph SLAM approach to refining the trajectory and closing loops. We evaluate the effectiveness of the system on using publicly available datasets with ground-truth data. The entire system is available for free and open-source online. [ABSTRACT FROM AUTHOR]

Purpose The most prominent example of scan matching algorithm is the Iterative Closest Point (ICP) algorithm. But the ICP algorithm and its variants excessively depend on the initial pose estimate between two scans. The purpose of this paper is to propose a scan matching algorithm, which is adaptable to big initial pose errors.Design/methodology/approach The environments are represented by flat units and upright units. The upright units are clustered to represent objects that the robot cannot cross over. The object cluster is further discretized to generate layered model consisting of cross-section ellipses. The layered model provides simplified features that facilitate an object recognition algorithm to discriminate among common objects in outdoor environments. A layered model graph is constructed with the recognized objects as nodes. Based on the similarity of sub-graphs in each scans, the layered model graph-based matching algorithm generates initial pose estimates and uses ICP to refine the scan matching results.Findings Experimental results indicate that the proposed algorithm can deal with bad initial pose estimates and increase the processing speed. Its computation time is short enough for real-time implementation in robotic applications in outdoor environments.Originality/value This paper proposes a bio-inspired scan matching algorithm for mobile robots based on layered model graph in outdoor environments. [ABSTRACT FROM AUTHOR]

In this paper, we present an algorithm for the exploration of an unknown graph by multiple robots, which is never worse than depth-first search with a single robot. On trees, we prove that the algorithm is optimal for two robots. For k robots, the algorithm has an optimal dependence on the size of the tree but not on its radius. We believe that the algorithm performs well on any tree, and this is substantiated by simulations. For trees with e edges and radius r, the exploration time is less than 2e/k+(1+(k/r))^k-1(2/k!)r^k-1 =(2e/k) + O((k+r)^k-1) (for r>k, \le (2e/k) + 2r^k-1), thereby improving a recent method with time O((e/\log k) + r) refid="ref2"/, and almost reaching the lower bound max ((2e/k), 2r). The model underlying undirected-graph exploration is a set of rooms connected by opaque passages; thus, the algorithm is appropriate for scenarios like indoor navigation or cave exploration. In this framework, communication can be realized by bookkeeping devices being dropped by the robots at explored vertices, the states of which are read and changed by further visiting robots. Simulations have been performed in both tree and graph explorations to corroborate the mathematical results. [ABSTRACT FROM AUTHOR]

Presents information on a study which used a backpropagation training algorithm with a scheme for the online tuning of a neural network tracking control system. Importance of robust stability and convergence to neural network systems; Methodology; Results; Conclusions.