Fig 1 - uploaded by Ioannis M. Rekleitis
Content may be subject to copyright.
(a) The placement and orientation of edge-map frames are determined by meet points in the GVG. (b) The resulting edge-maps are stored as individual, abstract structures.
Source publication
This paper presents a novel method of combining topological and feature-based mapping strategies to create a hierarchical approach to simultaneous localization and mapping (SLAM). More than simply running both processes in parallel, we use the topological mapping procedure to organize local feature-based methods. The result is an autonomous explora...
Context in source publication
Context 1
... of the GVG are either meet points, the set of points equidistant to three or more obstacles, or boundary points where the distance between two obstacles equals zero. These nodes are connected by edges which are paths of two-way equidis- tance, see the example in Figure 1. The definition of the nodes and edges automatically induce well defined control laws that allow a robot to trace an edge (either known or un- known a priori) and home onto a meet point. ...
Similar publications
Visual odometry in the field of computer vision and robotics is a well-known approach with which the position and orientation of an agent can be obtained using only images from a camera or multiple of them. In most traditional point feature-based visual odometry, one important assumption and also an ideal condition is that the scene remains static....
This paper describes a novel approach for focusing the search for victims of a rescue robot in an unknown environment. Such a rescue operation under time constraints is driven by three opposing factors, namely the need to explore unknown territory, the requirement to maintain robot localization performance as well as map quality in the face of unce...
Currently, some existing SLAM algorithms were widely applied in the field of mobile robots. However, most SLAM algorithms were tested either by simulation or using offline bags storing messages, which ignored other contributing factors influencing mapping including the changes of position and speed of the robot during it autonomous movement. The ai...
Laser range finder (LRF) or laser distance sensor (LDS), further referred to as LIDAR (light detection and ranging). LIDAR can obtain environmental point cloud data, while a robot can realize environmental sensing by adoption of the point cloud data generated and LIDAR-based SLAM (Simultaneous Localization And Mapping) algorithm. The precision of p...
A hopping rover is a robot that can move in low gravity planets by the characteristic motion called the hopping motion. For its autonomous explorations, the so-called SLAM (Simultaneous Localization and Mapping) is a basic function. SLAM is the combination of estimating the position of a robot and creating a map of an unknown environment. Most conv...
Citations
... AI in UAV can enhance its functioning many folds. Even they can be made to fly autonomously using different techniques like localization and mapping [14]. Various sensors can be used to localize and automate an UAV [15]. ...
Applications of unmanned aerial vehicles technology have shown a very big rise in the recent times. One of the main reasons is less cost, tropical deforestation and advancement in remote sensing technology. Artificial intelligence will become an integral part of unmanned aerial vehicles and can be used for various applications. Incorporation of such intelligence in a practical system is the need of hour. The aim of this paper is to embed artificial intelligence in drones using image processing. Intelligent drones are now the requirement of many fields right from courier delivery to defence, surveillance and rescue. Face-recognition system is proposed which is based on dataset creation, training and recognizer. Implementation of face-recognition system shows acceptable results. We have created an artificial intelligence which is capable of doing face recognition and incorporated it with an UAV.
... Modayil et al. [16] presented a combination of metric and topological SLAM, where local maps are constructed via metric information and the topology is utilized to solve loop closing. Similarly in [17] by Lisien et al., local feature-based maps are maintained as a low-level layer and a generalized Voronoi diagram serves as a high-level topological map organizing the lower level. An interesting work is presented in [18] from Wurm et al., where the standard RBPF metric-based algorithm is enhanced with a feature-based representation and the best representation for any moment is selected based on heuristics. ...
Rao-Blackwellized Particle Filters (RBPF) have been utilized to provide a solution to the SLAM problem. One of the main factors that cause RBPF failure is the potential particle impoverishment. Another popular approach to the SLAM problem are Scan Matching methods, whose good results require environments with lots of information, however fail in the lack thereof. To face these issues, in the current work techniques are presented to combine Rao-Blackwellized particle filters with a scan matching algorithm (CRSM SLAM). The particle filter maintains the correct hypothesis in environments lacking features and CRSM is employed in feature-rich environments while simultaneously reduces the particle filter dispersion. Since CRSM’s good performance is based on its high iteration frequency, a multi-threaded combination is presented which allows CRSM to operate while RBPF updates its particles. Additionally, a novel method utilizing topological information is proposed, in order to reduce the number of particle filter resamplings. Finally, we present methods to address anomalous situations where scan matching can not be performed and the vehicle displays behaviors not modeled by the kinematic model, causing the whole method to collapse. Numerous experiments are conducted to support the aforementioned methods’ advantages.
... In recent years, research interest in Unmanned Aerial Vehicles (UAVs) has increased due to its potential use for several applications such as: aerial photogrammetry [1], agriculture [2], habitat mapping [3] and military tasks [4]. This is possible in drones due to the ability to perform dangerous and repetitive tasks in remote and dangerous environments. ...
In this paper a fractional order proportional-derivative (FO-PD) control strategy is presented and applied to AR. Drone quadrotor system. The controller parameters are calculated based on specifying a certain gain crossover frequency, a phase margin and a robustness to gain variations. Its performance is compared against two other integer order controllers; i) Extended Prediction Self-Adaptive Control (EPSAC) approach to Model Predictive Control (MPC) ii) Integer order PD controller. The closed loop control simulations applied on the AR. Drone system indicate the proposed controller outperforms the integer order PD control. Additionally, the proposed controller has less complexity but similar performance as MPC based control.
... If corresponding images are available of the area, they can be overlain onto the point cloud surface. SLAM uses a set of algorithms or software to simultaneously locate itself while modeling in real-time (Haines, 2016). The SLAM system can be a laser-based system or visual based (using imagery). ...
Digital photogrammetry is becoming a more common method used for mapping geological and structural rock mass features in underground mining. The issue of capturing geological and structural data in inaccessible, unsupported areas of mines remains even when utilizing terrestrial photogrammetric methods; thus, geotechnical models of mines are left with incomplete datasets. Large unsupported underground voids, like stopes, have the potential to cause major failures, but by filling in the geotechnical data gaps in inaccessible areas, potential failures can be predicted through kinematic analysis of the area’s mapped discontinuities. Implementation of Unmanned Aerial Vehicles (UAVs) in underground mines and recent advances in obstacle detection systems have allowed for greater experimentation with photogrammetry conducted from a UAV platform in mines. For this study, a UAV-based underground photogrammetry system was developed to manually capture imagery in an inaccessible stope at Barrick Gold Corporation’s Golden Sunlight Mine (GSM) in Whitehall, Montana, to see whether or not the approach is a viable remote sensing technique for gathering georeferenced geotechnical data. Development of the system involved selecting an appropriate UAV platform, identifying a lighting system capable of providing adequate illumination, acquiring a sensor system that consistently avoids obstacles, and choosing the appropriate UAV camera (and its respective settings) for underground UAV-based imaging. In order to georeference the data collected in the inaccessible stope, paintballs were shot into the stope to create ground control points that were then surveyed in laser range detection. These paintball marks had to be in visual line-of-sight and visible in the imagery captured via UAV camera in order to georeferenced them. Using the imagery collected in the stope at GSM, models were constructed and structural features were mapped on those models. Bentley ContextCapture software was able to successfully construct a stope model from the video frame imagery collected via UAV in the stope, while ADAM Technology was not. Split-Engineering’s Split-FX and ADAM Technology were used separately to map the discontinuity planes present within the model. A comparison of underground discontinuity mapping was performed using the UAV-based photogrammetry captured in the stope and hand mapping data collected around the entrance to the stope. It was found that northeasterly striking discontinuity planes were identified using the digital mapping, but not in hand mapping. Using UAV-based photogrammetry for geotechnical data collection creates a quick and thorough mapping process with time-stamped imagery that can potentially create a safer mine. The lessons learned during this study may help guide future efforts using UAVs to capture geologic data and to help monitor stability in areas that are inaccessible.
... Grid-based maps supply discrete occupancy rates of destroyed areas, but the mobile robot's position cannot be located. 2 Hierarchical SLAM requires the local sub-maps to be statistically independent, which constrains data sharing among local metric sub-maps. Hierarchical local maps (HLMs) contain a set of sub-maps; at each time point, a new sub-map is added indicating and storing a mobile robot's position and covariance at that moment. ...
... To fuse higher topological maps with lower featurebased metric sub-maps, the generalized Voronoi graph (GVG) map method employs a global topological map to maintain consistency of local feature-based submaps. 2 A globally consistent map can be generated by merging local map patches according to a relative route graph; end points from the laser range finder (LRF) readings are transformed from the mobile robot. 13 Sensor noise drastically reduces the useful information of sensor readings; the noise is primarily generated by environmental factors such as surface texture and illumination, and by the measurement principle itself such as interference errors between ultrasonic sensors. ...
Perception-driven hierarchical simultaneous localization and mapping approach is proposed based on a distributed particle of percolator model adapted to search and rescue post-disaster environment. The proposed hierarchical analysis reduced the complexity; it can only highlight the interesting three-dimensional objects marked on the two-dimensional map, which adapted to urgent search and rescue post-disaster scenarios. Object identification, area coverage, and loop closure are the key issues that simultaneous localization and mapping faces when mobile robots detect and reach survivors in search and rescue collapsed environments. We developed hierarchical analysis using iterative extended Kalman filter method to improve the localization accuracy in simultaneous localization and mapping of mobile robots with an iteration process. An analytic hierarchical map-building mechanism efficiently constructs a globally consistent topological map using graph-based nodes and data association. An improved iterative extended Kalman filter procedure is used to estimate positions and the covariance matrix. The proposed algorithm incorporates a prediction step and a measurement update step, which performs the state vector time update in constant time interval and the correction of the state vector based on the new sensor measurement. The hierarchical mapping process is divided into three levels. The bottom level perceives the surroundings and collects input. The middle level receives and processes data from the bottom level. The upper level organizes the feature information from the previous levels to form a global, topologically consistent map. Simulations and experiments using the mobileRobot platform, Pioneer LX robot, and a crawler mobile robot validate the efficiency of this hierarchical simultaneous localization and mapping approach adapting to search and rescue post-disaster scenarios.
... A large number of studies have emerged in the literature on UAVs. Some examples of UAVs' applications can be found in precision agriculture [163], formation control of UGVs using an UAV [164], habitat mapping [165]. Modeling, identification and control of an UAV using on-board sensing are presented in [166]. ...
... Essentially, a drone is a flying robot. The aircraft may be remotely controlled or can fly autonomously through software-controlled flight plans in their embedded system working in conjunction with GPS [1,9,10]. ...
... For testing purpose and for material safety, tests have been performed in indoor environment. The major goal of these tests was to evaluate the correctness of adaptation capacity of adaptation of used algorithm [5], [6], [22]. During the tests, different factors have been varied: initial drone position, drone target coordinates (between 3 to 7 meters) and layout of possible obstacles. ...
... Modayil et al. [13] presented a combination of metric and topological SLAM, where local maps are constructed via metric information and the topology is employed to solve the loop closing problem. A similar idea is presented by Lisien et al. [14], where local feature-based maps are maintained as a low-level layer and a generalized Voronoi diagram serves as a high-level topological map organizing the lower level. ...
Two of the most predominant approaches regarding the SLAM problem are the Rao-Blackwellized particle filters and the Scan Matching algorithms, each approach presenting its own deficiencies. In particular, particle filters suffer from potential particle impoverishment, whereas lack of environmental features can cause scan matching methods to collapse. In the current paper a multi-threaded combination of Rao-Blackwellized particle filters with a scan matching algorithm (CRSM SLAM) aiming to overcome those defects, whilst exploiting each method's advantages is presented. CRSM is employed in feature-rich environments while concurrently reducing the particle filter dispersion, whilst the particle filter allows the maintenance of the correct hypothesis in environments with scarcity of information. Finally, a method to reduce the number of particle filter resamplings, employing topological information is proposed.
... 6 However, recent work has attempted to visually map the environment using low cost drones such as the Parot AR.Drone by using efficient algorithms. 7 Similar techniques have been used for quadcopter technology by other institutes, such as those described by researchers in California Institute of Technology. 8 Another pertinent example may be seen by the Computer Vision Group at Tehcnische Universitat Munchen (TUM). ...
The AR.Drone is a remote controlled quadcopter which is low cost, and readily available for consumers. Therefore it represents a simple test-bed on which control and vision research may be conducted. However, interfacing with the AR.Drone can be a challenge for new researchers as the AR.Drone's application programming interface (API) is built on low-level, bit-wise, C instructions. Therefore, this paper will demonstrate the use of an additional layer of abstraction on the AR.Drone’s API via the Robot Operating System (ROS). Using ROS, the construction of a high-level graphical user interface (GUI) will be demonstrated, with the explicit aim of assisting new researchers in developing simple control and vision algorithms to interface with the AR.Drone. The GUI, formally known as the Control and Vision Interface (CVI) is currently used to research and develop computer vision, simultaneous localisation and mapping (SLAM), and path planning algorithms by a number of postgraduate and undergraduate students at the school of Aeronautical, Mechanical, and Mechatronics Engineering (AMME) in The University of Sydney.
