This paper presents control and coordination algorithms for groups of vehicles. The focus is on autonomous vehicle networks performing distributed sensing tasks where each vehicle plays the role of a mobile tunable sensor. The paper proposes gradient descent algorithms for a class of utility functions which encode optimal coverage and sensing policies. The resulting closed-loop behavior is adaptive, distributed, asynchronous, and verifiably correct.

Coverage control for mobile sensing networks

Обнаружение транспортных средств на изображениях загородных шоссе на основе метода Single shot multibox Detector

This paper assumes a requirement for an unmanned multiunit satellite to be assembled in orbit. The requirement to be met is to bring the satellites together so tha t they do not collide but actually rendezvous. The equations of motion of the rendezvous satellite in a relative coordinate system are derived and used to compute a final injection velocity which would effect collision after a time r. The velocity is corrected periodically by a command guidance system and just before impact retrothrust is applied. A terminal infrared homing sj^stem is required to actually accomplish physical contact and joining of the satellites. The first satellite placed in orbit is the "control satellite" and controls all the satellites to be assembled and contains the ccmputer, command guidance equipment, precision orientation equipment, and other features necessary to effect rendezvous. The succeeding satellites contain a propulsion system, a rough at t i tude control system, and a command receiver plus whatever scientific equipment they carry to perform their basic mission. This paper presents the following:

Terminal Guidance System for Satellite Rendezvous

Filtering and smoothing methods are used to produce an accurate estimate of the state of a time-varying system based on multiple observational inputs (data). Interest in these methods has exploded in recent years, with numerous applications emerging in fields such as navigation, aerospace engineering, telecommunications, and medicine. This compact, informal introduction for graduate students and advanced undergraduates presents the current state-of-the-art filtering and smoothing methods in a unified Bayesian framework. Readers learn what non-linear Kalman filters and particle filters are, how they are related, and their relative advantages and disadvantages. They also discover how state-of-the-art Bayesian parameter estimation methods can be combined with state-of-the-art filtering and smoothing algorithms. The book’s practical and algorithmic approach assumes only modest mathematical prerequisites. Examples include MATLAB computations, and the numerous end-of-chapter exercises include computational assignments. MATLAB/GNU Octave source code is available for download at www.cambridge.org/sarkka, promoting hands-on work with the methods.

贝叶斯滤波与平滑 (Bayesian filtering and smoothing)

In the last few years, the deep learning (DL) computing paradigm has been deemed the Gold Standard in the machine learning (ML) community. Moreover, it has gradually become the most widely used computational approach in the field of ML, thus achieving outstanding results on several complex cognitive tasks, matching or even beating those provided by human performance. One of the benefits of DL is the ability to learn massive amounts of data. The DL field has grown fast in the last few years and it has been extensively used to successfully address a wide range of traditional applications. More importantly, DL has outperformed well-known ML techniques in many domains, e.g., cybersecurity, natural language processing, bioinformatics, robotics and control, and medical information processing, among many others. Despite it has been contributed several works reviewing the State-of-the-Art on DL, all of them only tackled one aspect of the DL, which leads to an overall lack of knowledge about it. Therefore, in this contribution, we propose using a more holistic approach in order to provide a more suitable starting point from which to develop a full understanding of DL. Specifically, this review attempts to provide a more comprehensive survey of the most important aspects of DL and including those enhancements recently added to the field. In particular, this paper outlines the importance of DL, presents the types of DL techniques and networks. It then presents convolutional neural networks (CNNs) which the most utilized DL network type and describes the development of CNNs architectures together with their main features, e.g., starting with the AlexNet network and closing with the High-Resolution network (HR.Net). Finally, we further present the challenges and suggested solutions to help researchers understand the existing research gaps. It is followed by a list of the major DL applications. Computational tools including FPGA, GPU, and CPU are summarized along with a description of their influence on DL. The paper ends with the evolution matrix, benchmark datasets, and summary and conclusion.

/pdf/review-of-deep-learning-concepts-cnn-architectures-hpbk62qzc2.pdf

Review of deep learning: concepts, CNN architectures, challenges, applications, future directions

This letter represents a new unscented information filtering algorithm for nonlinear estimation and multiple sensor information fusion. The proposed information fusion algorithm is derived by embedding the unscented transformation method used in the sigma point filter into the extended information filtering architecture. The new information filter achieves not only the accuracy and robustness of the sigma point filter but also the flexibility of the information filter for multiple sensor estimation. Performance comparison of the proposed filter with the extended information filter is demonstrated through a target-tracking simulation study.

Nonlinear Estimation and Multiple Sensor Fusion Using Unscented Information Filtering

The decision for the International Space Station (ISS) to maneuver to avoid a potential collision with another space object will be based on the probability of collision, P
C. The calculation of P
C requires the covariance of both objects at conjunction. It is well known that the covariance computed by US Space Command is optimistic (too small), especially at altitudes where atmospheric drag is the dominant perturbation, because its computation assumes there are no dynamic model errors. In this paper the effect of errors in the covariance on P
C and the sensitivity of P
C to the encounter geometry are investigated.

Probability of Collision Error Analysis

In recent years, demand has been increasing for target detection and tracking from aerial imagery via drones using onboard powered sensors and devices. We propose a very effective method for this application based on a deep learning framework. A state-of-the-art embedded hardware system empowers small flying robots to carry out the real-time onboard computation necessary for object tracking. Two types of embedded modules were developed: one was designed using a Jetson TX or AGX Xavier, and the other was based on an Intel Neural Compute Stick. These are suitable for real-time onboard computing power on small flying drones with limited space. A comparative analysis of current state-of-the-art deep learning-based multi-object detection algorithms was carried out utilizing the designated GPU-based embedded computing modules to obtain detailed metric data about frame rates, as well as the computation power. We also introduce an effective target tracking approach for moving objects. The algorithm for tracking moving objects is based on the extension of simple online and real-time tracking. It was developed by integrating a deep learning-based association metric approach with simple online and real-time tracking (Deep SORT), which uses a hypothesis tracking methodology with Kalman filtering and a deep learning-based association metric. In addition, a guidance system that tracks the target position using a GPU-based algorithm is introduced. Finally, we demonstrate the effectiveness of the proposed algorithms by real-time experiments with a small multi-rotor drone.

Deep Learning-Based Real-Time Multiple-Object Detection and Tracking from Aerial Imagery via a Flying Robot with GPU-Based Embedded Devices.

Deep Reinforcement Learning has led us to newer possibilities in solving complex control and navigation related tasks. The paper presents Deep Reinforcement Learning autonomous navigation and obstacle avoidance of self-driving cars, applied with Deep Q Network to a simulated car an urban environment. The approach uses two types of sensor data as input: camera sensor and laser sensor in front of the car. It also designs a cost-efficient high-speed car prototype capable of running the same algorithm in real-time. The design uses a camera and a Hokuyo Lidar sensor in the car front. It uses embedded GPU (Nvidia-TX2) for running deep-learning algorithms based on sensor inputs.

Driverless Car: Autonomous Driving Using Deep Reinforcement Learning in Urban Environment

This paper addresses the problem of how to optimally rendezvous with a set of satellites in geosynchronous orbit that have small inclinations. The method developed has application to two problems, servicing of geosynchronous satellites and the removal of dead satellites in geosynchronous orbit. It is shown that when inclination changes are required the optimal solution reduces to the solution of the traveling salesman problem. The method was applied to a set of non-operational geosynchronous satellites selected from the Space Object catalog. The method also directly identifies those satellites that are very costly (in fuel) to visit so that they can be removed from the list to be visited. NOMENCLATURE Ω right ascension I inclination λ longitude n mean motion h angular momentum v velocity T R g g , geosynchronous radius and orbit period τ jk jk N , time and number of orbits to travel from satellite j to satellite k Tm mission lifetime Ts servicing time for each satellite INTRODUCTION The servicing of satellites in orbit is a concept that is receiving more and more attention. If this could be achieved economically then the system lifetime could be extended and system costs reduced. The Hubble Space Telescope (HST) has been serviced to repair the solar panels, mirror and gyros. The cost of the HST made it worthwhile even though it was a very expensive mission in that it involved the use of astronauts and the Space Shuttle. For servicing to be economical satellites will have to be designed to be serviced. In addition, the servicing will have to be performed by robotic spacecraft; the use of man in space will make the costs prohibitive in most cases. The best economy will be achieved when numerous satellites can be serviced in one servicing mission. Since plane changes require a lot of fuel two scenarios in which numerous satellites could be serviced are constellations with many satellites in each plane, such as Iridium, and geosynchronous satellites. A few satellites could potentially be serviced in low Earth orbit utilizing differential nodal precession if all the satellites had approximately the same inclination. Even though there is a policy now that satellites are to be removed from the geosynchronous belt at end of life there are many non-operational satellites in geosynchronous orbit that are a hazard to other satellites. If there was a way to rendezvous with these satellites and attach a small engine to raise their orbit then over a period of time the geosynchronous belt could be made safer. The method developed in this paper applies to these two problems. The method determines the minimum fuel solution for visiting (rendezvousing with) a set of satellites in geosynchronous orbit with small inclinations. Since the fuel to rendezvous with a satellite in the same orbit is very small if sufficient time is allowed for the maneuver the method finds the order in which the satellites should be visited to minimize the fuel required for the plane changes. It is shown that the minimum fuel solution is proportional to the minimum distance path through the set of points that are the projections of the angular momentum AIAA/AAS Astrodynamics Specialist Conference and Exhibit 5-8 August 2002, Monterey, California AIAA 2002-4905 Copyright © 2002 by the author(s). Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. 2 American Institute of Aeronautics and Astronautics vectors on the equatorial plane. Thus, the minimum fuel (∆v) solution is the solution of the Traveling Salesman Problem (TSP). The fuel required for the in-plane maneuver is not minimized for several reasons. It is much smaller than the fuel required for the plane change and it can be obtained by vectoring the out of plane ∆v to obtain the necessary small in-plane component. Strategies for the rendezvous include equal time for each rendezvous or equal ∆v for the in-plane portion of the maneuver. Parameters in the problem are the total time and the servicing time at each satellite. The required ∆v is relatively insensitive to the servicing time until it becomes so large that only a short time is allowed for the transfer from one satellite to another. Then the in-plane ∆v begins to increase rapidly and dominate. ANALYSIS Consider a set, S, of N geosynchronous satellites where Si represents the ith satellite with orbital elements ei i i i T e e e = ( ) 1 2 6 , , , . K Find the order of visiting the satellites such that the total fuel ∆v ( ) to rendezvous (visit) is minimized. In addition, identify those satellites that should not be serviced because the fuel to service them is prohibitive. In this preliminary study we restrict the set S to consist of satellites in circular orbits and equal periods of 24 hours, small, but not necessarily zero, inclination and arbitrary right ascension. The circular orbit and 24 hour restrictions could be relaxed to consider near circular and near 24 hour satellites, such as the drifters. First we show that the finding the order of visiting each of the satellites that results in the minimum ∆v solution for the plane change portion of the ∆v to rendezvous with each of the N satellites in S is the solution of the Traveling Salesman Problem (TSP). Consider two satellites of the set S with inclination and right ascensions I I 1 1 2 2 , , Ω Ω ( ) ( ) and . Let X Y Z , , ( ) be an inertial coordinate system with Z along the polar axis, and X and Y define the equatorial plane. As shown in Figure 1 the coordinates of the projection of the angular momentum vector in the equatorial plane are h I j j j j j sin sin , cos , , Ω Ω − ( ) = 1 2. With equal angular momentum the distance d between the two points is

Deok Jin Lee

Papers

Nonlinear Estimation and Multiple Sensor Fusion Using Unscented Information Filtering

Probability of Collision Error Analysis

Deep Learning-Based Real-Time Multiple-Object Detection and Tracking from Aerial Imagery via a Flying Robot with GPU-Based Embedded Devices.

Driverless Car: Autonomous Driving Using Deep Reinforcement Learning in Urban Environment

Optimal Servicing of Geosynchronous Satellites