A GRASP-Based Approach for Planning UAV-Assisted Search and Rescue Missions

Search and Rescue (SAR) missions aim to search and provide first aid to persons in distress or danger. Due to the urgency of these situations, it is important to possess a system able to take fast action and effectively and efficiently utilise the available resources to conduct the mission. In addition, the potential complexity of the search such as the ruggedness of terrain or large size of the search region should be considered. Such issues can be tackled by using Unmanned Aerial Vehicles (UAVs) equipped with optical sensors. This can ensure the efficiency in terms of speed, coverage and flexibility required to conduct this type of time-sensitive missions. This paper centres on designing a fast solution approach for planning UAV-assisted SAR missions. The challenge is to cover an area where targets (people in distress after a hurricane or earthquake, lost vessels in sea, missing persons in mountainous area, etc.) can be potentially found with a variable likelihood. The search area is modelled using a scoring map to support the choice of the search sub-areas, where the scores represent the likelihood of finding a target. The goal of this paper is to propose a heuristic approach to automate the search process using scarce heterogeneous resources in the most efficient manner.

Visual tracking using multi-layer appearance approach

This paper investigated single target tracking of arbitrary objects. Tracking is a difficult problem due to a variety of challenges such as scale variation, motion, background clutter, illumination etc. To achieve better tracking performance under these severe conditions, this paper proposed covariance descriptor based on multi-layer instance search region. Our results show that the proposed approach significantly improves the performance in term of centre location error (in pixels) compared to covariance descriptor with using a fixed bounding box. From this work, it is believed that we have constructed a great solution in choosing best layer for this descriptor. This will be addressed in the next future work such as consider target motion during tracking.

WiFi Positioning in 3GPP Indoor Office with Modified Particle Swarm Optimization

With the start of the Fourth Industrial Revolution, Internet of Things (IoT), artificial intelligence (AI), and big data technologies are attracting global attention. AI can achieve fast computational speed, and big data makes it possible to store and use vast amounts of data. In addition, smartphones, which are IoT devices, are owned by most people. Based on these advantages, the above three technologies can be combined and effectively applied to navigation technology. In the case of an outdoor environment, global positioning system (GPS) technology has been developed to enable relatively accurate positioning of the user. However, due to the problem of radio wave loss because of many obstacles and walls, there are obvious limitations in applying GPS to indoor environments. Hence, we propose a method to increase the accuracy of user positioning in indoor environments using wireless-fidelity (Wi-Fi). The core technology of the proposed method is to limit the initial search region of the particle swarm optimization (PSO), an intelligent particle algorithm; doing so increases the probability that particles converge to the global optimum and shortens the convergence time of the algorithm. For this reason, the proposed method can achieve fast processing time and high accuracy. To limit the initial search region of the PSO, we first build an received signal strength indicator (RSSI) database for each sample point (SP) using a fingerprinting scheme. Then, a limited region is established through a fuzzy matching algorithm. Finally, the particles are randomly distributed within a limited region, and then the user’s location is positioned through a PSO. Simulation results confirm that the method proposed in this paper achieves the highest positioning accuracy, with an error of about 1 m when the SP interval is 3 m in an indoor environment.

A Fragmentation Region-based Skyline Computation Framework for a Group of Users

Skyline processing, an established preference evaluation technique, aims at discovering the best, most preferred objects, i.e. those that are not dominated by other objects, in satisfying the user’s preferences. In today’s society, due to the advancement of technology, ad-hoc meetings or impromptu gathering are becoming more and more common. Deciding on a suitable meeting point (object)for a group of people (users) to meet is not a straightforward task especially when these users are located at different places with distinct preferences. A place which is close by to the users might not provide the facilities/services that meet all the users’ preferences; while a place having the facilities/services that meet most of the users’ preferences might be too distant from these users. Although the skyline operator can be utilised to filter the dominated objects among the objects that fall in the region of interest of these users, computing the skylines for various groups of users in similar region would mean rescanning the objects of the region and repeating the process of pair wise comparisons among the objects which are undoubtedly unwise. On this account, this study presents a region-based skyline computation framework which attempts to resolve the above issues by fragmenting the search region of a group of users and utilising the past computed skyline results of the fragments. The skylines, which are the objects recommended to be visited by a group of users, are derived by analysing both the locations of the users, i.e. spatial attributes, as well as the spatial and non-spatial attributes of the objects. Several experiments have been conducted and the results show that our proposed framework outperforms the previous works with respect to CPU time.

Visual MAV Tracker with Adaptive Search Region

Tracking a micro aerial vehicle (MAV) is challenging because of its small size and swift motion. A new model was developed by combining compact and adaptive search region (SR). The model can accurately and robustly track MAVs with a fast computation speed. A compact SR, which is slightly larger than a target MAV, is less likely to include a distracting background than a large SR; thus, it can accurately track the MAV. Moreover, the compact SR reduces the computation time because tracking can be conducted with a relatively shallow network. An optimal SR to MAV size ratio was obtained in this study. However, this optimal compact SR causes frequent tracking failures in the presence of the dynamic MAV motion. An adaptive SR is proposed to address this problem; it adaptively changes the location and size of the SR based on the size, location, and velocity of the MAV in the SR. The compact SR without adaptive strategy tracks the MAV with an accuracy of 0.613 and a robustness of 0.086, whereas the compact and adaptive SR has an accuracy of 0.811 and a robustness of 1.0. Moreover, online tracking is accomplished within approximately 400 frames per second, which is significantly faster than the real-time speed.

Searching and Tracking an Unknown Number of Targets: A Learning-Based Method Enhanced with Maps Merging

Unmanned aerial vehicles (UAVs) have been widely used in search and rescue (SAR) missions due to their high flexibility. A key problem in SAR missions is to search and track moving targets in an area of interest. In this paper, we focus on the problem of Cooperative Multi-UAV Observation of Multiple Moving Targets (CMUOMMT). In contrast to the existing literature, we not only optimize the average observation rate of the discovered targets, but we also emphasize the fairness of the observation of the discovered targets and the continuous exploration of the undiscovered targets, under the assumption that the total number of targets is unknown. To achieve this objective, a deep reinforcement learning (DRL)-based method is proposed under the Partially Observable Markov Decision Process (POMDP) framework, where each UAV maintains four observation history maps, and maps from different UAVs within a communication range can be merged to enhance UAVs’ awareness of the environment. A deep convolutional neural network (CNN) is used to process the merged maps and generate the control commands to UAVs. The simulation results show that our policy can enable UAVs to balance between giving the discovered targets a fair observation and exploring the search region compared with other methods.

Improving the Optimality Verification and the Parallel Processing of the General Knapsack Linear Integer Problem

The chapter presents a new approach to improve the verification process of optimality for the general knapsack linear integer problem. The general knapsack linear integer problem is very difficult to solve. A solution for the general knapsack linear integer problem can be accurately estimated, but it can still be very difficult to verify optimality using the brach and bound related methods. In this chapter, a new objective function is generated that is also used as a more binding equality constraint. This generated equality constraint can be shown to significantly reduce the search region for the branch and bound-related algorithms. The verification process for optimality proposed in this chapter is easier than most of the available branch and bound-related approaches. In addition, the proposed approach is massively parallelizable allowing the use of the much needed independent parallel processing.

Generalized Multi-Scale Stochastic Reservoir Opportunity Index for enhanced well placement optimization under uncertainty in green and brownfields

Well placement planning is one of the challenging issues in any field development plan. Reservoir engineers always confront the problem that which point of the field should be drilled to achieve the highest recovery factor and/or maximum sweep efficiency. In this paper, we use Reservoir Opportunity Index (ROI) as a spatial measure of productivity potential for greenfields, which hybridizes the reservoir static properties, and for brownfields, ROI is replaced by Dynamic Measure (DM), which takes into account the current dynamic properties in addition to static properties. The purpose of using these criteria is to diminish the search region of optimization algorithms and as a consequence, reduce the computational time and cost of optimization, which are the main challenges in well placement optimization problems. However, considering the significant subsurface uncertainty, a probabilistic definition of ROI (SROI) or DM (SDM) is needed, since there exists an infinite number of possible distribution maps of static and/or dynamic properties. To build SROI or SDM maps, the k-means clustering technique is used to extract a limited number of characteristic realizations that can reasonably span the uncertainties. In addition, to determine the optimum number of clustered realizations, Higher-Order Singular Value Decomposition (HOSVD) method is applied which can also compress the data for large models in a lower-dimensional space. Additionally, we introduce the multiscale spatial density of ROI or DM (D2ROI and D2DM), which can distinguish between regions of high SROI (or SDM) in arbitrary neighborhood windows from the local SROI (or SDM) maxima with low values in the vicinity. Generally, we develop and implement a new systematic approach for well placement optimization for both green and brownfields on a synthetic reservoir model. This approach relies on the utilization of multi-scale maps of SROI and SDM to improve the initial guess for optimization algorithm. Narrowing down the search region for optimization algorithm can substantially speed up the convergence and hence the computational cost would be reduced by a factor of 4.