Abstract

In this paper, we present a practical vision based Simultaneous Localization and Mapping (SLAM) system for a highly dynamic environment. We adopt a multibody Structure from Motion (SfM) approach, which is the generalization of classical SfM to dynamic scenes with multiple rigidly moving objects. The proposed framework of multibody visual SLAM allows choosing between full 3D reconstruction or simply tracking of the moving objects, which adds flexibility to the system, for scenes containing non-rigid objects or objects having insufficient features for reconstruction. The solution demands a motion segmentation framework that can segment feature points belonging to different motions and maintain the segmentation with time. We propose a realtime incremental motion segmentation algorithm for this purpose. The motion segmentation is robust and is capable of segmenting difficult degenerate motions, where the moving objects is followed by a moving camera in the same direction. This robustness is attributed to the use of efficient geometric constraints and a probability framework which propagates the uncertainty in the system. The motion segmentation module is tightly coupled with feature tracking and visual SLAM, by exploring various feed-backs in between these modules. The integrated system can simultaneously perform realtime visual SLAM and tracking of multiple moving objects using only a single monocular camera.

Abstract

In this paper, we present a monocular camera based terrain classification scheme. The uniqueness of the proposed scheme is that it inherently incorporates spatial smoothness while segmenting a image, without requirement of post-processing smoothing methods. The algorithm is extremely fast because it is build on top of a Random Forest classifier. The baseline algorithm uses color, texture and their combination with classifiers such as SVM and Random Forests. We present comparison across features and classifiers. We further enhance the algorithm through a label transfer method. The efficacy of the proposed solution can be seen as we reach a low error rates on both our dataset and other publicly available datasets.

Abstract

In this paper we deduce the evolution of a four wheeled active suspension rover from a five wheeled passive suspension rover. The aim of this paper is to design a suspension mechanism which utilizes the advantages of both passive suspension and active suspension rover. Both the design considered here are simpler than the existing suspension mechanisms in the sense that the number of links as wells as the number of joints have been significantly reduced without compromising the climbing capability of the rover. We first analyze the kinematics of the five wheeled rover and its motion pattern while climbing an obstacle and try to deduce the same motion pattern and capability in the four wheeled rover. Both the suspension mechanism consists of two planar closed kinematic chains on each side of the rover. We also deduce the control strategy for the active suspension rover wherein only two actuators are used to control the internal configuration of the rover. To the best of author's knowledge this is the minimum number of actuators required to control the internal configuration of a active suspension while operating on a fully 3D rough terrain. Extensive uneven terrain simulations are performed for both 5-wheeled and 4-wheeled rover and a comparative analysis has been done on maximum coefficient of friction and torque requirements.

Abstract

The ability to detect, and track multiple moving objects like person and other robots, is an important prerequisite for mobile robots working in dynamic indoor environments. We approach this problem by detecting independently moving objects in image sequence from a monocular camera mounted on a robot. We use multi-view geometric constraints to classify a pixel as moving or static. The first constraint, we use, is the epipolar constraint which requires images of static points to lie on the corresponding epipolar lines in subsequent images. In the second constraint, we use the knowledge of the robot motion to estimate a bound in the position of image pixel along the epipolar line. This is capable of detecting moving objects followed by a moving camera in the same direction, a so-called degenerate configuration where the epipolar constraint fails. To classify the moving pixels robustly, a Bayesian framework is used to assign a probability that the pixel is stationary or dynamic based on the above geometric properties and the probabilities are updated when the pixels are tracked in subsequent images. The same framework also accounts for the error in estimation of camera motion. Successful and repeatable detection and pursuit of people and other moving objects in realtime with a monocular camera mounted on the Pioneer 3DX, in a cluttered environment confirms the efficacy of the method.

Abstract

In environments with identical features, the global localization of a robot, might result in multiple hypotheses of its location. If the situation is extrapolated to multiple robots, it results in multiple hypotheses for multiple robots. The localization is facilitated if the robots are actively guided towards locations where it can use other robots as well as obstacles to localize itself. This paper aims at presenting a learning technique for the above process of active localization of multiple robots by co-operation. An MDP framework is used for learning the task, over a semi-decentralized team of robots hereby maintaining a bounded complexity as opposed to various multi-agent learning techniques, which scale exponentially with the increase in the number of robots.

Abstract

We present efficient measurement models for localization in a feature based EKF SLAM framework. Both points and segments form the features, points include corners formed by intersection of wall like segments. The point features are described by its coordinates, while the segment feature is represented by the angle made by the normal to the segment from the global origin with the abscissa called the normal angle or N-angle for short. The measurement equation involves measuring the distance and bearing to the point feature and only bearing to the line feature. The distance measurement to the segment is intentionally kept out of the measurement equation due to its inefficacy in correcting the robot and landmarks state. This arises due to very large differences in the predicted and observed distances even for modest measurement errors when the robot is not very near the segment. Hence for this reason the segment feature is represented only by its N-angle devoid of distance since such a representation results in better state correction. The number of computations resulting from the covariance matrix updates is also less than a representation that includes both N-distance and N-angle.

Abstract

We present a method for extracting ground and other planes from a single non rotating laser mounted on a slow moving car used for on-road driving. A laser scan is decomposed into linear clusters. Corresponding clusters from subsequent scans are merged to form planes. The ground plane is identified based on the current vehicle height and the variance in height of the planes. Once these seed planes are identified future scan points either get associated with these planes or result in formation of new planes. Scan points that do not belong to any of the plane are left as such in the representation. Since the robustness of the method is contingent on how a single scan is decomposed into linear clusters, we compare the quality of the terrain representation due to three such clustering methods, one by iterative end point fit, other by adaptive breakpoint detection and thirdly the current method based on adaptive cosine similarity.

Abstract

We present a fast multi-robotic exploration methodology for 2D and 3D terrains. An asynchronous exploration strategy is introduced which shows significant improvements over the existing synchronous ones. A per-time visibility metric is being utilized by the algorithm. The metric allots the same weight for points for next view whose visibility over time ratios are equal. The outcome of this is that while the number of points visited to explore a terrain is nearly the same as other popular metrics found in literature, the time length of the paths are smaller in this case resulting in reduced time exploration. The results have been verified through extensive simulations in 2D and 3D. In 2D multiple robots explore unknown terrains that are office like, cluttered, corridor like and various combinations of these. In 3D we consider the case of multiple UAVs exploring a terrain represented as height fields. We introduce a way for calculating expected visibilities and a way of incorporating explored features in the per-time metric. The maximum height of the UAV at each location is governed by the so called exposure surface, beneath which the UAVs are constrained to fly. We also show performance gain of the present metric over others in experiments on a Pioneer 3DX robot.

Abstract

Global localization algorithms involve a search over all possible poses of the robot that can be typically over a large space in huge maps. Essentially it involves computing a posterior by seeing how probable are the obtained sensor readings at each of the discretized states in a map. Instead in this paper by reverse projecting the sensor readings from the obstacle boundaries onto the surroundings, a solution is obtained by searching over the space of obstacle boundaries than by a search in the discretized pose space. That this search over obstacle boundaries is considerably less if the ratio of free space to boundary space in a map is high is straightforward. However we also show theoretically that even when the boundary space exceeds the free space the computations due to the current method does not exceed those due to the popular Markov and Correlation based approaches to global localization. The comparative advantages are well documented in simulation section of the paper. The approach is able to consistently localize a laser equipped robot in our lab.

Abstract

We present a method for coordinating multi-robotic/multi-agent traffic control at intersections. The robotic agents (RA) move guided by a potential field along the lanes. At the intersections an intersection agent (IA) controls the flow of traffic by assigning priorities to the agents that are about to enter the intersection. The priorities are computed based on the density of RA in a lane and the flow rate of traffic in those lanes. The RAs integrate these assigned priorities into their potential field computations. The modified potential field computations help the RAs to move through the intersection avoiding collisions. An elegant mixed autonomy scheme is thereby achieved where the IAs decide upon the priorities at the intersection while the low level collision avoidance maneuvers are left with the individual RAs. This scheme preserves the distributed nature and the autonomy of potential field maneuvers while simultaneously balancing the computation load between the IA and RAs. We compare this method with a method where the RAs navigate the intersection without a superior direction from the IAs through priorities or when IAs direct the RAs based on priorities computed on a first come first served basis. We show performance gain over both these methods in simulations.