Intelligent transportation systems are a subset of system of systems that combine information from autonomous vehicles, road infrastructure, and other systems in order to improve the safety and efficiency of travel on roadways. The performance of an intelligent transportation system is limited by the transmission distance and delay of critical messages between components of the system of systems. Roadside units are road infrastructure designed to improve both of these limiting factors through their superior transmission capabilities and the ability to tunnel information to other roadside units over long distances. The placement of roadside units is critical to the their ability to assist the intelligent transportation system, with authors applying a number of different methods for determining the optimal positions. The goal of this work is to prove the ability of neural networks to solve this problem, which up to this point has not been attempted. A convolutional neural network is presented which uses images of the road network to determine the optimal placement of a roadside unit. The network is trained using data obtained through OpenStreetMaps, and the results demonstrate the ability of neural networks to determine the optimal placement of a roadside unit.

Smart cities is a broad field that has been gaining traction in research in recent years. One particular topic that has been the subject of research for a number of years but is beginning to appear on the consumer market is intelligent transportation systems (ITS). As ITS become more common on roadways, infrastructure such as Roadside Units (RSUs) will be needed to support and improve them. RSUs are used to pass information between vehicles which can then be used to improve the safety and efficiency of the system. The placement of RSUs can greatly effect the efficiency of the infrastructure, which is crucial when the deployment of such systems can be expensive. Recent research has focused on a broad range of variables and methods to optimize the deployment of RSUs. The goal of this paper is to discuss recent research trends in the field and present a brief overview of a number of recent papers on the topic.

The use of deep reinforcement learning (DRL) as a framework for training a mobile robot to perform optimal navigation in an unfamiliar environment is a suitable choice for implementing AI with real-time robotic systems. In this study, the environment and surrounding obstacles of an Ackermann-steered UGV are reconstructed into a virtual setting for training the UGV to centrally learn the optimal route (guidance actions to be taken at any given state) towards a desired goal position using Multi-Agent Virtual Exploration in Deep Q-Learning (MVEDQL) for various model configurations. The trained model policies are to be transferred to a physical vehicle and compared based on their individual effectiveness for performing autonomous waypoint navigation. Prior to incorporating the learned model with the physical UGV for testing, this paper outlines the development of a GUI application to provide an interface for remotely deploying the vehicle and a virtual reality framework reconstruction of the training environment to assist safely testing the system using the reinforcement learning model.

This paper presents the application of an Unmanned Ground Vehicle(UGV) in a testbed of heterogeneous autonomous systems. The UGV comprises of three essential factors of navigation: A Mapping System, Localization system, and an Obstacle Avoidance System. The UGV will avoid obstacles using Light Detection and Ranging (LIDAR) technology compatible with a controller to prevent it from colliding with objects while reaching a goal. While the UGV is exploring the territory, it will use mapping techniques for visual reference. Robot Operating System manages the sensors into individual topics, which communicate together to link data for navigation.

The aptitude to identify the emotional states of others and response to exposed emotions is an important aspect of human social intelligence. Robots are expected to be prevalent in society to assist humans in various tasks. Human-robot interaction (HRI) is of critical importance in the assistive robotics sector. Smart digital assistants and assistive robots fail quite often when a request is not well defined verbally. When the assistant fails to provide services as desired, the person may exhibit an emotional response such as anger or frustration through expressions in their face and voice. It is critical that robots understand not only the language, but also human psychology. A novel affection-based perception architecture for cooperative HRIs is studied in this paper, where the agent is expected to recognize human emotional states, thus encourages a natural bonding between the human and the robotic artifact. We propose a method to close the loop using measured emotions to grade HRIs. This metric will be used as a reward mechanism to adjust the assistant's behavior adaptively. Emotion levels from users are detected through vision and speech inputs processed by deep neural networks (NNs). Negative emotions exhibit a change in performance until the user is satisfied.

Robotic systems have been gathering more data as the field advances, allowing them to complete much more difficult tasks. This data has the potential to cause issues in the system once it reaches too large a size, and therefore must be preprocessed to manageable levels. This paper proposes a method to remove redundant data points from a data set based on a model formed from radial basis function (RBF) approximation. An optimization problem based off the Euclidean distance between the complete and reduced model with an additional least absolute shrinkage and selection operator (LASSO) component is formulated. A solution to the problem is then solved using simulated annealing. Finally, two simulations are set up to assess the performance of the algorithm.

Researchers have extensively explored indoor localization in recent years. Robotic applications often require precise positioning, which is difficult when access to a GPS is limited or compromised. Many of the existing approaches require either prior knowledge of the local environment and fixed landmarks or complex and expensive hardware to achieve the necessary degree of accuracy. In this article, we examine the use of the HTC virtual reality hardware along with our novel approach, VIVEPOSE, to perform indoor robotic localization. We then compare the accuracy of our proposed indoor localization system to current approaches that use traditional odometry sensor-based localization. Finally, we present and demonstrate a leader-follower approach using VIVEPOSE and show how the HTC Vive tracker can be successful for indoor localization for a robotics application.

The field of swarm robotics has continued to grow in recent years, branching into new fields that require more complex algorithms to solve and faster response times to control. Fields such as cloud and fog computing have been used to counter this first issue, allowing agents to use a network to access systems with much higher computational abilities, but this comes at the cost of increased latency. New architectures such as the Lambda and Kappa architectures have been developed to remedy this issue, allowing for cloud computers to be used while not sacrificing response times. This paper focuses on comparing three different architectures to assess their performance on a simulated robotic swam to solve the mobile sensor optimization problem. These three architectures are a traditional robot, the Lambda architecture, and the Kappa architecture. These systems are compared based off their latency and the runtimes of different algorithms in each implementation.

This paper focuses on data fusion, which is fundamental to one of the most important modules in any autonomous system: perception. Over the past decade, there has been a surge in the usage of smart/autonomous mobility systems. Such systems can be used in various areas of life like safe mobility for the disabled, senior citizens, and so on and are dependent on accurate sensor information in order to function optimally. This information may be from a single sensor or a suite of sensors with the same or different modalities. We review various types of sensors, their data, and the need for fusion of the data with each other to output the best data for the task at hand, which in this case is autonomous navigation. In order to obtain such accurate data, we need to have optimal technology to read the sensor data, process the data, eliminate or at least reduce the noise and then use the data for the required tasks. We present a survey of the current data processing techniques that implement data fusion using different sensors like LiDAR that use light scan technology, stereo/depth cameras, Red Green Blue monocular (RGB) and Time-of-flight (TOF) cameras that use optical technology and review the efficiency of using fused data from multiple sensors rather than a single sensor in autonomous navigation tasks like mapping, obstacle detection, and avoidance or localization. This survey will provide sensor information to researchers who intend to accomplish the task of motion control of a robot and detail the use of LiDAR and cameras to accomplish robot navigation.

Cloud computing is now a global standard computing topology and has been widely studied for many years. Less frequently researched is the use of cloud and edge computing to optimize the performance of a system as a whole. One important aspect of cloud and edge computing is managing the placement of the applications in the network system so as to minimize each application's runtime, given the resources of system's devices and the capabilities of the system's network. The properties of containerized applications now make this possible. The process of containerization creates a lightweight, mobile, packaged application for each of the algorithms in a system. These applications can then be deployed easily and quickly on any layer of the cloud and edge computing architectures. In this research, a fuzzy placement control system is designed to place applications on the cloud-edge model. As verified by simulation, the fuzzy placement control system proposed reduces the total runtime of the applications by placing applications in an efficient location.

In this article a widely studied consensus algorithm for processing sensor data of multimissile systems is considered. A novelty of algorithms to reach a consensus given time-delay constraints is introduced. In previous work, the author proposed an algorithm addressing the same problem. The primary goal of this article is to evaluate the performance of the proposed algorithm. The performance was evaluated in Monte Carlo runs in a target tracking scenario. A mission with multimissile trying to intercept a single flying threat is assumed in this scenario. The interceptors are using nonhomogeneous sensors to measure the location of the threat continuously. Each missile computes its state prediction and shares it with its neighboring missile only. However, the shared information is applied to arbitrary nonuniform time delay. The entire group of missiles must reach a consensus on the target's location. The performance is evaluated in terms of overall estimation error, conflict resolution, and time to reach a consensus. Different scenarios are also simulated to examine the effectiveness of the number of missiles performing observation, the delay value, and sensor uncertainty in the overall system stability. Moreover, the tradeoffs of using the algorithm are identified.

As the demand for mobile autonomous systems increases across various industries, fault diagnostic systems will need to become more intelligent and robust. In this paper we propose a distributed Long Short-Term Memory (LSTM)- based ensemble learning architecture for learning highly nonlinear, temporal fault classification boundaries for an Unmanned Ground Vehicle (UGV). The main goal of the architecture is to reduce classification bias by ensembling LSTM models as well as achieving near-real time processing time. This is done by parallelizing the deep learning models on Amazon Web Services (AWS) cloud instances via Apache Kafka, a real-time data pipelining infrastructure. An experiment is conducted on a UGV subjected to dislocated suspension faults and results showing the effectiveness of the approach are shown.

A system is as good as its sensors and in turn a sensor is only as good as the data it measures. Accurate, optimal sensor data can be used in autonomous systems applications such as environment mapping, obstacle detection and avoidance, and similar applications. In order to obtain such accurate data we need to have optimal technology to read the sensor data, process the data, eliminate the noise and then utilize it. As part of this paper, we present a survey of the current data processing techniques that implement data fusion using various sensors like LiDAR, stereo/depth cameras and RGB monocular cameras and we mention the implementations or usage of this fused data in tasks like obstacle detection and avoidance or localization etc. In future, we plan on implementing a state-of-art fusion system, on an intelligent wheelchair, controlled by a human thought, without intervention of any of the user’s motor skills. We propose an efficient and fast algorithm that senses and learns the projection from the camera space to the LiDAR space and outputs camera data in the form of LiDAR detection (distance and angle) and a multi-sensor and multi-modal detection system that fuses both the camera and LiDAR detections to obtain better accuracy and robustness.

Artificial intelligence (AI) has been an issue in robotics, since AI is based on iterative algorithms. In general, simulations of physical models are used to show the outcome of learning algorithms or show proof of concepts. Since models are generated based on parameter estimations of training data, it is crucial to iterate a significant amount of times in order to model an accurate classification function. Thus, it would take a substantial amount of time for a robot to generate such a function. In this research, an implementation of reinforced learning will be applied on a unmanned ground vehicle (UGV) learning simulated model that will be translated into a physical UGV using Robot Operating System (ROS) to test performance of the given model.

The following topics are dealt with: systems engineering; Internet of Things; mobile robots; security of data; control engineering computing; learning (artificial intelligence); road vehicles; data visualisation; transportation; health care.

The field of robotics research is continuously expanding at an ever-increasing rate. So much so, that as a systems' complexity grows, so too does the amount of possible points of failure. In recent years, these systems have been integrated together to create systems of systems, dramatically increasing the fragility of these networked systems, also known as a swarm. This paper presents a method for abstracting the fault of a networked control system, namely a system of mobile robots, into general feature sets and producing the capability of predicting the present fault as well as the compensation thereof.

Virtual reality (VR) has become a very popular gaming platform in recent years, but has not been used as much in the research domain. Several technologies used to create virtual environments can also be used to assist in research. For example, recent developments in VR allow for easier integration of humans in the loop for control and manipulation of complex, multimodal system of systems. In this paper we show some of the methods in which a virtually emulated system can be integrated into an artificial intelligence (AI) based system of systems. We demonstrate how this integration allows the reseacher to more easily understand and analyse the inter-system dynamics, and to improve performance using post-simulation visualization, real-time simulation interaction, hardware emulation, and physical object tracking. The results of using virtual reality as a simulation environment shows the usefulness of the tool to system of systems researchers.

With improvements in reinforcement learning algorithms, and the demand to implement these algorithms on real systems, the use of a simulator as an intermediate stage is essential to save time, material and financial resources. The lack of particular features in a unified simulator for applications to autonomous cars and robotics, encouraged this research, which produced a simulator capable of simulating multiple car like objects, in either one or several arenas (environments). Being a lightweight application, multiple instances of the simulator can run at the same time, only constrained by the available computational resources.