Berat A. Erol

Formation control implementation using Kobuki TurtleBots and Parrot Bebop drone

Formation control of a collection of vehicles is a topic that has generated a lot of interest in the research community. This interest primarily stems from the increased performance and robustness that is provided by a swarm of agents as compared to an individual member. Formation control can be achieved through many approaches. The approach used by this paper is based on a leader-follower premise. A network of agents can be controlled by assigning a leader for each agent in the formation. The group as a whole will be capable of following either a Virtual Leader (VL) or an agent within the group. The algorithm applied to a test-bed consisting of three Kobuki TurtleBot2 robots. Each Turtlebot2 is programmed to follow a pre-defined virtual point in the formation. The test space is monitored by a Parrot Bebop drone hovering overhead that identifies agents uniquely through image processing techniques. The agents can then move in the test space, based on the leaders position, while maintaining a formation.

A deep vision landmark framework for robot navigation

Robot navigation requires specific techniques for guiding a mobile robot to a desired destination. In general, a desired path is required in an environment described by different terrain and a set of distinct objects, such as obstacles and particular landmarks. In this paper, a new approach for autonomous navigation is presented using machine learning techniques such as Convolutional Neural Network to identify markers or objects from images and Robot Operating System and Object Position Discovery system to orient to the marker, calculate the distance and navigate towards these markers using depth camera.

Proposed testbed for the modeling and control of a system of autonomous vehicles

Large scale multi-agent systems are very important in todays world because of their varying uses. The Center for Testing, Evaluation and Control of Heterogeneous Large scale Autonomous Vehicles (TECHLAV) has been tasked to conduct research on Large Scale Autonomous Systems of Vehicles (LSASV). This paper focuses on the proposed testbed system that will help model the large scale system out in the field for Modeling, Analysis and Control tasks for LSASV (MACLAV). The system will use a team of UGVs, UAVs and AUVs to navigate, interact and complete missions through an unknown area as a cohesive unit. A small private cloud provides a computational backbone to the system.

Cloud-based control and vSLAM through cooperative Mapping and Localization

Simultaneous Localization and Mapping (SLAM) is one of the most widely popular and applied methods designed for more accurate localization and navigation operations. Our experiments showed that vision based mapping helps agents navigate in unknown environments using feature based mapping and localization. Instead of using a classical monocular camera as a vision source, we have decided to use RGB-D (Red, Green, Blue, Depth) cameras for better feature detection, 3D mapping, and localization. This is due to the fact that the RGB-D camera returns depth data as well as the normal RGB data. Moreover, we have applied this method on multiple robots using the concept of cooperative SLAM. This paper illustrates our current research findings and proposes a new architecture based on gathered data from RGB-D cameras, which are the Microsoft Kinect and the ASUS Xtion Pro for 3D mapping and localization.

Application of Big Data Analytics via Cloud Computing

Advances in sensor technology, the Internet of things (IoT), social networking, wireless communications and huge collection of data from years have all contributed to a new field of study Big Data is discussed in this paper. The System of Systems (SoS) integrates independently operating, non-homogeneous systems to achieve a higher goal than the sum of the parts. Recently, management of data has become strenuous and SoS helps in solving the problems and providing solutions, with the new approaches in Data Analytics. Data Analytics uses both statistical and cloud computing using machine learning or computational intelligence to reduce the size of Big Data to a manageable size to extract information, build a knowledge base using the derived data, and eventually develop a nonparametric model for the Big Data. In this research, the approaches towards the cloud environment for Data Analytics is discussed which is one of the key application areas of Big Data. Through this analysis and survey, we provide recommendations for the research community on future directions on providing data-based decisions for cloud-supported Big Data computing and analytic solutions.

Software interface design for home-based assistive multi-robot system

In many assistive robotic systems, the interface to the user is simply a tablet computer or a monitor attached to a single robot. Missing from approaches are the system extensibility made possible with a tablet computer and a division of work between multiple agents. In this paper we present the design for a software interface to connect users to an assistive robot system for the disabled and elderly. The system is comprised of heterogeneous low-cost assistive robots, a home management portal and a cloud computing backend. The system is designed with the premise that all components do not need to be present for the system to function, but it will be improved when expanded by addition of robots and expanded computing capabilities. This paper focuses on developing the interfaces necessary to connect the user to these systems in a simple and easy to comprehend manner for the target user population.

Improved Deep Neural Network Object Tracking System for Applications in Home Robotics

Robotic navigation in GPS-denied environments requires case specific approaches for controlling a mobile robot to any desired destinations. In general, a nominal path is created in an environment described by a set of distinct objects, in other words such obstacles and landmarks. Intelligent voice assistants or digital assistance devices are increasing their importance in today’s smart home. Especially, by the help of fast-growing Internet of Things (IoT) applications. These devices are amassing an ever-growing list of features such as controlling states of connected smart devices, recording tasks, and responding to queries. Assistive robots are the perfect complement to smart voice assistants for providing physical manipulation. A request made by a person can be assigned to the assistive robot by the voice assistant. In this chapter, a new approach for autonomous navigation is presented using pattern recognition and machine learning techniques such as Convolutional Neural Networks to identify markers or objects from images and videos. Computational intelligence techniques are implemented along with Robot Operating System and object positioning to navigate towards these objects and markers by using RGB-depth camera. Multiple potential matching objects detected by the robot with deep neural network object detectors will be displayed on a screen installed on the assistive robot to improve and evaluate Human-Robot Interaction (HRI).

Improved route optimization for autonomous ground vehicle navigation

Mobile robot navigation is a very important exercise in all robotic application from a domestic household cleaner to highly dangerous life threatening situations. Path planning is the main issue related to navigation. It is very important for an autonomous mobile vehicle to navigate properly without any collision or unsafe conditions in its environment. Path planning in mobile robots must ensure the optimal route with least cost and collision free path. In this paper, an algorithm to reduce the path searching process by half of the traditional methods is proposed.