The Internet of Cooperative Agents Architecture (X-IoCA) for Robots, Hybrid Sensor Networks, and MEC Centers in Complex Environments: A Search and Rescue Case Study.

Cloud robotics and advanced communications can foster a step-change in cooperative robots and hybrid wireless sensor networks (H-WSN) for demanding environments (e.g., disaster response, mining, demolition, and nuclear sites) by enabling the timely sharing of data and computational resources between robot and human teams. However, the operational complexity of such multi-agent systems requires defining effective architectures, coping with implementation details, and testing in realistic deployments. This article proposes X-IoCA, an Internet of robotic things (IoRT) and communication architecture consisting of a hybrid and heterogeneous network of wireless transceivers (H2WTN), based on LoRa and BLE technologies, and a robot operating system (ROS) network. The IoRT is connected to a feedback information system (FIS) distributed among multi-access edge computing (MEC) centers. Furthermore, we present SAR-IoCA, an implementation of the architecture for search and rescue (SAR) integrated into a 5G network. The FIS for this application consists of an SAR-FIS (including a path planner for UGVs considering risks detected by a LoRa H-WSN) and an ROS-FIS (for real-time monitoring and processing of information published throughout the ROS network). Moreover, we discuss lessons learned from using SAR-IoCA in a realistic exercise where three UGVs, a UAV, and responders collaborated to rescue victims from a tunnel accessible through rough terrain.

Development and Implementation of a Hybrid Wireless Sensor Network of Low Power and Long Range for Urban Environments.

The urban population, worldwide, is growing exponentially and with it the demand for information on pollution levels, vehicle traffic, or available parking, giving rise to citizens connected to their environment. This article presents an experimental long range (LoRa) and low power consumption network, with a combination of static and mobile wireless sensors (hybrid architecture) to tune and validate concentrator placement, to obtain a large coverage in an urban environment. A mobile node has been used, carrying a gateway and various sensors. The Activation By Personalization (ABP) mode has been used, justified for urban applications requiring multicasting. This allows to compare the coverage of each static gateway, being able to make practical decisions about its location. With this methodology, it has been possible to provide service to the city of Malaga, through a single concentrator node. The information acquired is synchronized in an external database, to monitor the data in real time, being able to geolocate the dataframes through web mapping services. This work presents the development and implementation of a hybrid wireless sensor network of long range and low power, configured and tuned to achieve efficient performance in a mid-size city, and tested in experiments in a real urban environment.

Integration of a Mobile Node into a Hybrid Wireless Sensor Network for Urban Environments.

Robots, or in general, intelligent vehicles, require large amounts of data to adapt their behavior to the environment and achieve their goals. When their missions take place in large areas, using additional information to that gathered by the onboard sensors frequently offers a more efficient solution of the problem. The emergence of Cyber-Physical Systems and Cloud computing allows this approach, but integration of sensory information, and its effective availability for the robots or vehicles is challenging. This paper addresses the development and implementation of a modular mobile node of a Wireless Sensor Network (WSN), designed to be mounted onboard vehicles, and capable of using different sensors according to mission needs. The mobile node is integrated with an existing static network, transforming it into a Hybrid Wireless Sensor Network (H-WSN), and adding flexibility and range to it. The integration is achieved without the need for multi-hop routing. A database holds the data acquired by both mobile and static nodes, allowing access in real-time to the gathered information. A Human⁻Machine Interface (HMI) presents this information to users. Finally, the system is tested in real urban scenarios in a use-case of measurement of gas levels.

Analysis of 3D Scan Measurement Distribution with Application to a Multi-Beam Lidar on a Rotating Platform.

Multi-beam lidar (MBL) rangefinders are becoming increasingly compact, light, and accessible 3D sensors, but they offer limited vertical resolution and field of view. The addition of a degree-of-freedom to build a rotating multi-beam lidar (RMBL) has the potential to become a common solution for affordable rapid full-3D high resolution scans. However, the overlapping of multiple-beams caused by rotation yields scanning patterns that are more complex than in rotating single beam lidar (RSBL). In this paper, we propose a simulation-based methodology to analyze 3D scanning patterns which is applied to investigate the scan measurement distribution produced by the RMBL configuration. With this purpose, novel contributions include: (i) the adaption of a recent spherical reformulation of Ripley's K function to assess 3D sensor data distribution on a hollow sphere simulation; (ii) a comparison, both qualitative and quantitative, between scan patterns produced by an ideal RMBL based on a Velodyne VLP-16 (Puck) and those of other 3D scan alternatives (i.e., rotating 2D lidar and MBL); and (iii) a new RMBL implementation consisting of a portable tilting platform for VLP-16 scanners, which is presented as a case study for measurement distribution analysis as well as for the discussion of actual scans from representative environments. Results indicate that despite the particular sampling patterns given by a RMBL, its homogeneity even improves that of an equivalent RSBL.

Voxel-Based Neighborhood for Spatial Shape Pattern Classification of Lidar Point Clouds with Supervised Learning.

Improving the effectiveness of spatial shape features classification from 3D lidar data is very relevant because it is largely used as a fundamental step towards higher level scene understanding challenges of autonomous vehicles and terrestrial robots. In this sense, computing neighborhood for points in dense scans becomes a costly process for both training and classification. This paper proposes a new general framework for implementing and comparing different supervised learning classifiers with a simple voxel-based neighborhood computation where points in each non-overlapping voxel in a regular grid are assigned to the same class by considering features within a support region defined by the voxel itself. The contribution provides offline training and online classification procedures as well as five alternative feature vector definitions based on principal component analysis for scatter, tubular and planar shapes. Moreover, the feasibility of this approach is evaluated by implementing a neural network (NN) method previously proposed by the authors as well as three other supervised learning classifiers found in scene processing methods: support vector machines (SVM), Gaussian processes (GP), and Gaussian mixture models (GMM). A comparative performance analysis is presented using real point clouds from both natural and urban environments and two different 3D rangefinders (a tilting Hokuyo UTM-30LX and a Riegl). Classification performance metrics and processing time measurements confirm the benefits of the NN classifier and the feasibility of voxel-based neighborhood.

Principal components analysis competitive learning.

We present a new neural model that extends the classical competitive learning by performing a principal components analysis (PCA) at each neuron. This model represents an improvement with respect to known local PCA methods, because it is not needed to present the entire data set to the network on each computing step. This allows a fast execution while retaining the dimensionality-reduction properties of the PCA. Furthermore, every neuron is able to modify its behavior to adapt to the local dimensionality of the input distribution. Hence, our model has a dimensionality estimation capability. The experimental results we present show the dimensionality-reduction capabilities of the model with multisensor images.

A principal components analysis self-organizing map.

We propose a new self-organizing neural model that performs principal components analysis. It is also related to the adaptive subspace self-organizing map (ASSOM) network, but its training equations are simpler. Experimental results are reported, which show that the new model has better performance than the ASSOM network.