Autonomous Scene Exploration for Robotics: A Conditional Random View-Sampling and Evaluation Using a Voxel-Sorting Mechanism for Efficient Ray Casting

Carrying out the task of the exploration of a scene by an autonomous robot entails a set of complex skills, such as the ability to create and update a representation of the scene, the knowledge of the regions of the scene which are yet unexplored, the ability to estimate the most efficient point of view from the perspective of an explorer agent and, finally, the ability to physically move the system to the selected Next Best View (NBV). This paper proposes an autonomous exploration system that makes use of a dual OcTree representation to encode the regions in the scene which are occupied, free, and unknown. The NBV is estimated through a discrete approach that samples and evaluates a set of view hypotheses that are created by a conditioned random process which ensures that the views have some chance of adding novel information to the scene. The algorithm uses ray-casting defined according to the characteristics of the RGB-D sensor, and a mechanism that sorts the voxels to be tested in a way that considerably speeds up the assessment. The sampled view that is estimated to provide the largest amount of novel information is selected, and the system moves to that location, where a new exploration step begins. The exploration session is terminated when there are no more unknown regions in the scene or when those that exist cannot be observed by the system. The experimental setup consisted of a robotic manipulator with an RGB-D sensor assembled on its end-effector, all managed by a Robot Operating System (ROS) based architecture. The manipulator provides movement, while the sensor collects information about the scene. Experimental results span over three test scenarios designed to evaluate the performance of the proposed system. In particular, the exploration performance of the proposed system is compared against that of human subjects. Results show that the proposed approach is able to carry out the exploration of a scene, even when it starts from scratch, building up knowledge as the exploration progresses. Furthermore, in these experiments, the system was able to complete the exploration of the scene in less time when compared to human subjects.

A New Volumetric CNN for 3D Object Classification Based on Joint Multiscale Feature and Subvolume Supervised Learning Approaches

The advancement of low-cost RGB-D and LiDAR three-dimensional (3D) sensors has permitted the obtainment of the 3D model easier in real-time. However, making intricate 3D features is crucial for the advancement of 3D object classifications. The existing volumetric voxel-based CNN approaches have achieved remarkable progress, but they generate huge computational overhead that limits the extraction of global features at higher resolutions of 3D objects. In this paper, a low-cost 3D volumetric deep convolutional neural network is proposed for 3D object classification based on joint multiscale hierarchical and subvolume supervised learning strategies. Our proposed deep neural network inputs 3D data, which are preprocessed by implementing memory-efficient octree representation, and we propose to limit the full layer octree depth to a certain level based on the predefined input volume resolution for storing high-precision contour features. Multiscale features are concatenated from multilevel octree depths inside the network, aiming to adaptively generate high-level global features. The strategy of the subvolume supervision approach is to train the network on subparts of the 3D object in order to learn local features. Our framework has been evaluated with two publicly available 3D repositories. Experimental results demonstrate the effectiveness of our proposed method where the classification accuracy is improved in comparison to existing volumetric approaches, and the memory consumption ratio and run-time are significantly reduced.

3-D mechanical analysis of complex reservoirs: a novel mesh-free approach

SUMMARY Building geomechanical models for induced seismicity in complex reservoirs poses a major challenge, in particular if many faults need to be included. We developed a novel way of calculating induced stress changes and associated seismic moment response for structurally complex reservoirs with tens to hundreds of faults. Our specific target was to improve the predictive capability of stress evolution along multiple faults, and to use the calculations to enhance physics-based understanding of the reservoir seismicity. Our methodology deploys a mesh-free numerical and analytical approach for both the stress calculation and the seismic moment calculation. We introduce a high-performance computational method for high-resolution induced Coulomb stress changes along faults, based on a Green's function for the stress response to a nucleus of strain. One key ingredient is the deployment of an octree representation and calculation scheme for the nuclei of strain, based on the topology and spatial variability of the mesh of the reservoir flow model. Once the induced stress changes are evaluated along multiple faults, we calculate potential seismic moment release in a fault system supposing an initial stress field. The capability of the approach, dubbed as MACRIS (Mechanical Analysis of Complex Reservoirs for Induced Seismicity) is proven through comparisons with finite element models. Computational performance and suitability for probabilistic assessment of seismic hazards are demonstrated though the use of the complex, heavily faulted Gullfaks field.

INDOOR A* PATHFINDING THROUGH AN OCTREE REPRESENTATION OF A POINT CLOUD

There is a growing demand of 3D indoor pathfinding applications. Researched in the field of robotics during the last decades of the 20th century, these methods focussed on 2D navigation. Nowadays we would like to have the ability to help people navigate inside buildings or send a drone inside a building when this is too dangerous for people. What these examples have in common is that an object with a certain geometry needs to find an optimal collision free path between a start and goal point. <br><br> This paper presents a new workflow for pathfinding through an octree representation of a point cloud. We applied the following steps: 1) the point cloud is processed so it fits best in an octree; 2) during the octree generation the interior empty nodes are filtered and further processed; 3) for each interior empty node the distance to the closest occupied node directly under it is computed; 4) a network graph is computed for all empty nodes; 5) the A* pathfinding algorithm is conducted. <br><br> This workflow takes into account the connectivity for each node to all possible neighbours (face, edge and vertex and all sizes). Besides, a collision avoidance system is pre-processed in two steps: first, the clearance of each empty node is computed, and then the maximal crossing value between two empty neighbouring nodes is computed. The clearance is used to select interior empty nodes of appropriate size and the maximal crossing value is used to filter the network graph. Finally, both these datasets are used in A* pathfinding.

An Improved Three-Dimensional Target Octonion Representation

It is an efficient way to represent three-dimensional objects by octree.The traditional structure of pointer -based octree representation has several shortcomings,such as requiring large memory,missing relationship between two nodes,etc.Based on analyzing the space Iayout and the configuration of octree,this paper presents an improved octree for 3D representation.From the experimental results for 3D reconstruction of medical images,we can see the proposed method is superior to the traditional method in terms of the storing structure and visiting way,etc.

Scheduling safe movement of air traffic in crowded air spaces

This paper considers the problem of generating conflict-free movement schedules for a set of vehicles that are operating simultaneously in a common airspace. In both civilian air traffic management and military air campaign planning contexts, it is crucial that the movements of different vehicles be coordinated so as to avoid collisions and near misses. Our approach starts from a view of airspace management as a 4D resource allocation problem, where the space in which vehicles must maneuver is itself managed as a capacitated resource. We introduce a linear octree representation of airspace capacity to index vector-based vehicle routes and efficiently detect regions of potential conflict. Generalizing the notion of contention-based search heuristics, we next define a scheduling algorithm that proceeds by first solving a relaxed version of the problem to construct a spatial capacity profile (represented as an octree), and then using spatio-temporal regions where demand exceeds capacity to make conflict-avoiding vehicle routing and scheduling decisions. We illustrate the utility of this basic representation and search algorithm in two ways. First, to demonstrate the overall viability of the approach, we present experimental results using data representing a realistically sized air campaign planning domain. Second, we define a more abstract notion of ‘encounter set’, which tolerates some amount of conflict on the assumption that on-board deconfliction processes can take appropriate avoidance maneuvers at execution time, and show that generation of this more abstract form of predictive guidance can be obtained without loss in computational efficiency.