Large Complex Environments Research Articles

Thanks to repetition, dustbathing detection can be automated combining accelerometry and wavelet analysis

AbstractBirds from at least a dozen orders engage in dustbathing, including Galliformes. Dustbathing is generally considered a behavioural need for poultry. It involves a precise and orderly sequence of movements repeated over time. The most characteristic movement involves tossing the dust with the wings and undulating the body beneath the dust shower. Thus, repetitive changes in body position during dustbathing could be automatically detected through data processing of body‐mounted accelerometer recordings. The approach was tested in 13 adult male Japanese quail (Coturnix japonica) fitted with a body mounted triaxial accelerometer. Behaviour was video‐recorded for at least 6 h. Observations showed that when the animal lies on its left‐ or right‐side during dustbathing, the lateral (swaying) component of the acceleration vector adopts values of +1 or −1, respectively. Analysis shows that the bird repeats these shifts in body position every 25–60 s. The wavelet analysis (i.e. complex Morlet continuous wavelet transform (CWT)) detected this oscillatory behaviour within the time series as higher power values. This characteristic was used to automate the detection of dustbathing events, for which a threshold value for the maximum power value estimated was established for the corresponding range of scales between 25 and 60 s. The overall general accuracy of this classification method for dustbathing detection was 80%, with individual variations falling within the range of 66%–100%. Finally, an example of the potential of this method in the study of temporal dynamics, such as daily rhythms of dustbathing, is provided. Our results show that combining accelerometry and wavelet analysis could be useful for the assessment of intra‐ and inter‐individual variability in dustbathing dynamics over long‐term studies, even within large complex environments, such as natural habitats or breeding facilities. Moreover, this approach could open doors for future in‐depth studies exploring the relationship between dustbathing and poultry welfare.

Multi-UAV Path Planning for Wireless Data Harvesting With Deep Reinforcement Learning

Harvesting data from distributed Internet of Things (IoT) devices with multiple autonomous unmanned aerial vehicles (UAVs) is a challenging problem requiring flexible path planning methods. We propose a multi-agent reinforcement learning (MARL) approach that, in contrast to previous work, can adapt to profound changes in the scenario parameters defining the data harvesting mission, such as the number of deployed UAVs, number, position and data amount of IoT devices, or the maximum flying time, without the need to perform expensive recomputations or relearn control policies. We formulate the path planning problem for a cooperative, non-communicating, and homogeneous team of UAVs tasked with maximizing collected data from distributed IoT sensor nodes subject to flying time and collision avoidance constraints. The path planning problem is translated into a decentralized partially observable Markov decision process (Dec-POMDP), which we solve through a deep reinforcement learning (DRL) approach, approximating the optimal UAV control policy without prior knowledge of the challenging wireless channel characteristics in dense urban environments. By exploiting a combination of centered global and local map representations of the environment that are fed into convolutional layers of the agents, we show that our proposed network architecture enables the agents to cooperate effectively by carefully dividing the data collection task among themselves, adapt to large complex environments and state spaces, and make movement decisions that balance data collection goals, flight-time efficiency, and navigation constraints. Finally, learning a control policy that generalizes over the scenario parameter space enables us to analyze the influence of individual parameters on collection performance and provide some intuition about system-level benefits.

Hierarchical Path-Planning for Mobile Robots Using a Skeletonization-Informed Rapidly Exploring Random Tree*

An efficient, hierarchical, two-dimensional (2D) path-planning method for large complex environments is presented in this paper. For mobile robots moving in 2D environments, conventional path-planning algorithms employ single-layered maps; the proposed approach engages in hierarchical inter- and intra-regional searches. A navigable graph of an environment is constructed using segmented local grid maps and safe junction nodes. An inter-regional path is obtained using the navigable graph and a graph-search algorithm. A skeletonization-informed rapidly exploring random tree* (SIRRT*) efficiently computes converged intra-regional paths for each map segment. The sampling process of the proposed hierarchical path-planning algorithm is locally conducted only in the start and goal regions, whereas the conventional path-planning should process the sampling over the entire environment. The entire path from the start position to the goal position can be achieved more quickly and more robustly using the hierarchical approach than the conventional single-layered method. The performance of the hierarchical path-planning is analyzed using a publicly available benchmark environment.

Data collection methods for studying pedestrian behaviour: A systematic review

Collecting pedestrian behaviour data is vital to understand pedestrian behaviour. This systematic review of 145 studies aims to determine the capability of contemporary data collection methods in collecting different pedestrian behavioural data, identify research gaps and discuss the possibilities of using new technologies to study pedestrian behaviour. The review finds that there is an imbalance in the number of studies that feature various aspects of pedestrian behaviour, most importantly (1) pedestrian behaviour in large complex scenarios, and (2) pedestrian behaviour during new types of high-risk situations. Additionally, three issues are identified regarding current pedestrian behaviour studies, namely (3) little comprehensive data sets featuring multi-dimensional behaviour data simultaneously, (4) generalizability of most collected data sets is limited, and (5) costs of pedestrian behaviour experiments are relatively high. A set of new technologies offers opportunities to overcome some of these limitations. This review identifies three types of technologies that can become a valuable addition to pedestrian behaviour research methods, namely (1) applying VR experiments to study pedestrian behaviour in the environments that are difficult or cannot be mimicked in real-life, repeat experiments to determine the impact of factors on pedestrian behaviour and collect more accurate behavioural data to understand the decision-making process of pedestrian behaviour deeply, (2) applying large-scale crowd monitoring to study pedestrian movements in large complex environments and incident situations, and (3) utilising the Internet of Things to track pedestrian movements at various locations that are difficult to investigate at the moment.

Coverage Path Planning for Decomposition Reconfigurable Grid-Maps Using Deep Reinforcement Learning Based Travelling Salesman Problem

Optimizing the coverage path planning (CPP) in robotics has become essential to accomplish efficient coverage applications. This work presents a novel approach to solve the CPP problem in large complex environments based on the Travelling Salesman Problem (TSP) and Deep Reinforcement Learning (DRL) leveraging the grid-based maps. The proposed algorithm applies the cellular decomposition methods to decompose the environment and generate the coverage path by recursively solving each decomposed cell formulated as TSP. A solution to TSP is determined by training Recurrent Neural Network (RNN) with Long Short Term Memory (LSTM) layers using Reinforcement Learning (RL). We validated the proposed method by systematically benchmarked with other conventional methods in terms of path length, execution time, and overlapping rate under four different map layouts with various obstacle density. The results depict that the proposed method outperforms all considered parameters than the conventional schemes. Moreover, simulation experiments demonstrate that the proposed approach is scalable to the larger grid-maps and guarantees complete coverage with efficiently generated coverage paths.

Optimized Wireless Channel Characterization in Large Complex Environments by Hybrid Ray Launching-Collaborative Filtering Approach

Simulation techniques based on deterministic methods such as ray tracing and ray launching are widely used to perform radioplanning tasks. However, the quality of the simulations depends on the number of rays and the angular resolution. The computational cost of these simulations in high definition prevents their use in complex environment, and their low-definition counterparts are used instead. In this letter, we propose a technique based on collaborative filtering to lessen the poor-quality problems of low-definition simulations. We show that our approach obtains results very similar to those of high definition in much less time. Also, we compare our approach to other well-known techniques and show that it performs better in terms of accuracy and precision. The use of combined deterministic/collaborative filtering techniques allows the estimation of radioplanning tasks in large, complex scenarios with a potentially large amount of transceivers.

Cooperative Exploration and Networking While Preserving Collision Avoidance.

Monitoring of large complex environments, such as underwater environments, is an important task in surveillance. An information (sensor) network can be built to achieve the task. To build an information network in an unknown workspace, we use multiple robots deploying information nodes. While robots build the network, they localize themselves as well as deployed nodes in the global coordinate system. Our multirobot networking strategy is as follows: each robot iteratively visits a frontier, which borders an unsensed area, until all areas are explored. As multiple robots explore the workspace, a robot must avoid colliding with another robot as well as with an obstacle. Hence, we introduce collision avoidance control laws and integrate the control laws with our cooperative networking strategy. Using MATLAB simulations, we verify the scalability and effectiveness of both our networking strategy and the collision avoidance control laws.

Reconstructing gas distribution maps via an adaptive sparse regularization algorithm

In this paper, we present an algorithm to be used by an inspection robot to produce a gas distribution map and localize gas sources in a large complex environment. The robot, equipped with a remote gas sensor, measures the total absorption of a tuned laser beam and returns integral gas concentrations. A mathematical formulation of such measurement facility is a sequence of Radon transforms, which is a typical ill-posed problem. To tackle the ill-posedness, we develop a new regularization method based on the sparse representation property of gas sources and the adaptive finite-element method. In practice, only a discrete model can be applied, and the quality of the gas distribution map depends on a detailed 3-D world model that allows us to accurately localize the robot and estimate the paths of the laser beam. In this work, using the positivity of measurements and the process of concentration, we estimate the lower and upper bounds of measurements and the exact continuous model (mapping from gas distribution to measurements), and then create a more accurate discrete model of the continuous tomography problem. Based on adaptive sparse regularization, we introduce a new algorithm that gives us not only a solution map but also a mesh map. The solution map more accurately locates gas sources, and the mesh map provides the real gas distribution map. Moreover, the error estimation of the proposed model is discussed. Numerical tests for both the synthetic problem and practical problem are given to show the efficiency and feasibility of the proposed algorithm.

Navigation and steering for autonomous virtual humans.

The ever-increasing applicability of interactive virtual worlds in industry and academia has given rise to the need for robust, versatile autonomous virtual humans to inject life into these environments. There are two fundamental problems that must be addressed to produce functional, purposeful autonomous populaces: (1)Navigation: finding a collision-free global path from an agent's start position to its target in large complex environments, and (2) Steering: moving an agent along the path while avoiding static and dynamic threats such as other agents. In this review, we survey the large body of contributions in steering and navigation for autonomous agents in dynamic virtual worlds. We describe the benefits and limitations of different proposed solutions and identify potential future research directions to meet the needs for the next generation of interactive virtual world applications. WIREs Cogn Sci 2013, 4:263-272. doi: 10.1002/wcs.1223 For further resources related to this article, please visit the WIREs website.

Novel Information Matrix Sparsification Approach for Practical Implementation of Simultaneous Localization and Mapping

Simultaneous localization and mapping (SLAM) is a fundamental issue in mobile robotics because it is the basis of higher-level tasks of robots. Recently, more and more research has been proposed that aims to enhance the efficiency of SLAM solutions from the viewpoint of the information matrix. This paper presents a novel, efficient SLAM approach by using the characters of the information matrix. Our approach eliminates many of the elements in the information matrix while maintaining the consistency. The large complex environment simulation, as well as outdoor car park experiment verifies the validity of our approach. The proposed sparsification method provides an efficient way to obtain a consistent estimation with provable upper bounds of sparsification errors.

7.1.4 Coping With System Integration Challenges in Large Complex Environments

AbstractThe increasing scope of systems integration poses many challenges for the system integrators. This paper contends that in conventional projects the Systems Integration phase is systematically underestimated. During this phase all unexpected and unforeseen problems surface, where most problems are cross‐organizational. When the project size increases with more suppliers, more users, more enterprise processes, and more (intelligent) functionality, then the system integration phase gets tremendously more difficult.In this paper we will bring order to the integration process and discuss an approach and best practices to cope with systems integration in conventional projects. Such level of integration approach is a prerequisite to deal with all added integration complexity in systems with a scope of (intelligent) enterprises.

Hospital Design and Wayfinding

A videotape simulation of the entrance drive to the new University of Michigan Hospital was used to test the effect on turning behavior of constructing a parking deck entrance directly off the hospital's drop-off circle. One hundred randomly sampled visitors were shown this videotape and asked where they would turn, if they were coming alone to visit a patient and needed to park their car. There were several turn-off areas along the entrance drive before the parking deck turn-off. Each turn-off had a sign listing its respective functions and directed drivers to continue straight ahead for parking. Half the visitors saw a videotape in which there was an entrance to the deck from the drop-off circle, and half saw a tape in which the entrance from the drop-off circle was absent. Each visitor saw two scenarios, one having the drop-off circle crowded with cars, and one uncrowded. The results of the study showed that the presence of the entrance to the deck from the drop-off circle did make a significant difference in reported turning behavior. The results of this study had a direct impact on the subsequent design decision. In addition, the results of this study are meaningful beyond the data themselves. The study used an inexpensive, simple, and highly imageable research technique, yet one rarely used in environment-behavior research. It was a "quick turn-around" research effortthat resolved a specific design dilemma. The study also demonstrates that signs alone are not sufficient to guide wayfinding behavior in large complex environments like hospitals.