This PDF file contains the front matter associated with SPIE Proceedings Volume 10195, including the Title Page, Copyright information, Table of Contents, Introduction (if any), and Conference Committee listing.

The Army Research Laboratory’s Robotics Collaborative Technology Alliance is a program intended to change robots from tools that soldiers use into teammates alongside which soldiers can work. This requires the integration of fundamental and applied research in robotic perception, intelligence, manipulation, mobility, and human-robot interaction. In this paper, we present the results of assessments conducted in 2016 to evaluate the capabilities of a new robot, the Robotic Manipulator (RoMan), and of a cognitive architecture (ACT-R). The RoMan platform was evaluated on its ability to conduct a search and grasp task under a variety of conditions. Specifically, it was required to search for and recognize a gas can placed on the floor, and then pick it up. The RoMan showed the potential to be a good platform for autonomous manipulation, but the autonomy used in these experiments will require improvement to make full use of the platform’s capabilities. The cognitive architecture was evaluated as to how well it could learn to select an appropriate set of features for a classification task. The task was to classify emotions that had been encoded using the Facial Action Coding System, with ACT-R learning to select the most effective set of features for correct classification. ACT-R leaned rules which required it to observe about half of the available features to make a decision, and the subsequent decisions had an accuracy ranging from 76% to 93% (depending on the emotion).

We propose a method to for robot planning using deep learning to integrate object detection and natural language understanding. This is different from other techniques, such the RCTA World Model,¹ which explicitly defines the interfaces between each module. These boundaries simplify the design and testing of complex robotic systems, but also introduce constraints that may reduce overall system performance. For example, perception tasks generate large amounts of data, but much of it is discarded to simplify interpretation by higher level tasks, e.g., a 3D object becomes a point in space, or a distribution over classifications is reduced to its mode. Further, errors in the robot’s overal task are not back-propagated to lower level tasks, and therefore, these tasks never adapt themselves to improve robot performance. We intend to address this by using a deep learning framework to replace these interfaces with learned interfaces that select what data is shared between modules and allow for error back-propagation that could adapt each module to the robot’s task. We will do this in a simplified system that accepts aerial orthographic images and simple commands and generates paths to achieve the command. Paths are learned from expert example via inverse optimal control. In time, we hope to evolve this simplified architecture towards something more complex and practical.

Legged robots are capable of navigating rough terrain, but have traditionally been restricted to slow speeds. New robots combine the power density necessary for rapid motions with increasingly sophisticated leg designs. Developing controllers that effectively coordinate these high-DOF legs to generate fast, agile motions is challenging. In this paper we examine a pair of control approaches to generate high-speed trotting for the direct-drive quadruped robot Minitaur. We first show that optimization of a redesigned feed-forward trajectory improves the robot’s running speed by 45%, from 1.52m/s to 1.93m/s. We then utilize a monopod version of Minitaur’s 5-bar leg to directly compare this control approach to a dynamic, model-based strategy. We find gaits with the optimized trajectory are able to achieve speeds up to 2.44m/s, but the model-based dynamic controller is able to find gaits that are more robust to parameter changes, nearly as fast, and up to 70% more efficient.

As robots are deployed in more dynamic and uncertain environments, the ability to recover from tip-over events is critical. Previously, a framework for generating quasi-static self-righting solutions for generic robots was developed. This paper extends that framework to include the use of inertial appendage methods. It begins by reviewing the basic framework and discussing how it may be extended to incorporate dynamic solutions. It then discusses the generation of appendage momentum in the presence of ground reaction forces by utilizing the zero moment point concept. This concept is further extended to controlling the momentum transfer between the appendage and the body such that a desired tip-over event results. After initiating the tip-over event, the motion of the appendage may further be controlled to reduce the energy of the impact to land within the basin of attraction, or to increase the energy to intentionally land outside that basin and continue the roll. Four strategies based on this methodology are introduced, permuting appendage acceleration or deceleration and whether or not the appendage is involved in the resulting ground contact. The strategies are compared based on three optimization metrics: energy required to induce tipping, collision energy, and stability margin. Finally, the proposed methods are validated on a physical robot, demonstrating the improvement to its rightability as compared with quasi-static solutions.

Mobile robots operating in the field need to be aware of the terrain and use this information to select the proper motion planning modality. For example, in very benign terrain it might be desirable to navigate to a goal following the shortest path. However, on challenging terrains, such maneuvers may be prohibitively expensive in terms of energy consumption. This paper summarizes field experiment results corresponding to a comparison between distance optimal and energy optimal motion planning for a skid-steered robot on different terrains. The results show that on average there is substantive energy savings associated with energy optimal motion planning.

Motion planning for legged machines such as RHex-type robots is far less developed than motion planning for wheeled vehicles. One of the main reasons for this is the lack of kinematic and dynamic models for such platforms. Physics based models are difficult to develop for legged robots due to the difficulty of modeling the robot-terrain interaction and their overall complexity. This paper presents a data driven approach in developing a kinematic model for the X-RHex Lite (XRL) platform. The methodology utilizes a feed-forward neural network to relate gait parameters to vehicle velocities.

Many aspects of decision making processes for autonomous systems involve human subjective information in some form. Methods for informing decision making processes with human information are needed to inform probabilistic information used in an autonomous system. This can provide better decisions and permit a UAV to more quickly and efficiently complete tasks. Specifically we use possibility theory to represent the subjective information and apply possibilistic conditioning of the probability distribution. A simulation platform was developed to evaluate approaches to using possibilistic inputs and showed that is was feasible to make effective usage of such information.

Autonomous navigation is a desired capability for mobile robotic platforms, such as UAVs and UGVs. Efficient autonomous navigation, especially in unknown environments, requires accurate localization based on online location recognition. Traditional monocular camera-based location recognition methods can suffer from accuracy, speed and robustness issues. In this paper, a novel online location recognition (OLR) algorithm based on fast and efficient interestpoint detection and feature-based keypoint matching is presented. The OLR incrementally constructs a database of visited locations and robustly recognizes revisited locations, even in complex and cluttered scenes. The OLR capability is quantitatively evaluated using a mobile robot setup in a multi-room office building environment. We further present and validate two applications based on the proposed OLR algorithm. In the first application, the OLR algorithm is integrated with conventional SLAM and an efficient trajectory correction algorithm for drift-free trajectory estimation (DTE) and qualitatively evaluated in the same environment. In the second application, the OLR algorithm is integrated with an exploration strategy for efficient autonomous navigation (EAN) and evaluated using a UAV model in a simulated warehouse environment. Our experiments demonstrate the accuracy and efficiency of the proposed location recognition algorithm and its utility for drift-free trajectory estimation and efficient autonomous navigation.

As unmanned systems (UMS) proliferate for security and defense applications, autonomous control system capabilities that enable them to perform tactical operations are of increasing interest. These operations, in which UMS must match or exceed the performance and speed of people or manned assets, even in the presence of dynamic mission objectives and unpredictable adversary behavior, are well beyond the capability of even the most advanced control systems demonstrated to date. In this paper we deconstruct the tactical autonomy problem, identify the key technical challenges, and place them into context with the autonomy taxonomy produced by the US Department of Defense’s Autonomy Community of Interest. We argue that two key capabilities beyond the state of the art are required to enable an initial fieldable capability: rapid abstract perception in appropriate environments, and tactical reasoning. We summarize our work to date in tactical reasoning, and present initial results from a new research program focused on abstract perception in tactical environments. This approach seeks to apply semantic labels to a broad set of objects via three core thrusts. First, we use physics-based multi-sensor fusion to enable generalization from imperfect and limited training data. Second, we pursue methods to optimize sensor perspective to improve object segmentation, mapping and, ultimately, classification. Finally, we assess the potential impact of using sensors that have not traditionally been used by UMS to perceive their environment, for example hyperspectral imagers, on the ability to identify objects. Our technical approach and initial results are presented.

Autonomous Air Vehicles (AUV) are used for survey, patrol and exploration purposes. Command and control of such vehicles is usually done by a human operator through use of joystick and visual camera. Most of the perception, target detection, and maneuvering is also done by a human. This approach does not allow for autonomy to evolve much further. Additionally, path planning algorithms are limited in scope and do not account for real-time events. A proposed fly eye antenna array system, gives an opportunity to scan its surroundings and detect multiple targets. It consists of angularly – spaced directional and overlapping antennas with wide – area coverage. Each directional antenna is coupled with front end circuit, and a digital processor. Signals coming from such an antenna array records information about position, direction, target proximity with good penetration through scattered media [1]. Object detection using radio frequency signals from such sensors is challenging. The object reflects electromagnetic signals, coming from multitude of the fly-eye antennas. Based on the correlative values obtained from the sensors, a determination is made of the location, shape, and size of the object. Such patterns are recorded and used for convolutional neural network training purposes. The reflected and positional information of the radar sensors provides valuable information to the pattern system: 9 sensor covering 360 C view, with a correlative signal strength matrix. The objective of using signals from antenna array of signals with together with artificially labeled data makes a system intelligent. This approach is much simpler and we think it will have better generalization and performance then other RF signal processing approaches.

In 2010 the ONR Code 30 Irregular and Expeditionary Warfare Department began a long-term strategic investment in ground system autonomy focused on supporting our Expeditionary forces, namely the United States Marine Corps (USMC). The Marine Corps' mission and challenges are unique and as such require unique capabilities of the autonomous systems they will employ. For the past six years ONR Code 30 has been executing the Multi-role Autonomous Ground Vehicle (MAGV) program which has focused on developing core technologies to enable the operation of autonomous ground vehicles in expeditionary environments. The key overarching tenets of the program have been to develop a ground autonomy system that is affordable, can operate in those expeditionary (off-route) environments, and which will be the technological foundation for future DoD programs both in S and T and acquisition. Those three key tenets have driven many of the investment decisions of the program including a focus on low-cost sensors and computation (e.g., vision systems), congested environment motion planning, operation under degraded GPS conditions, multi-vehicle coordination, and using a rapid development and evolutionary systems engineering approach with rigorous and consistent performance evaluations. This paper describes the goals and objectives of the ONR 30 ground autonomy program and provides an overview of the technical accomplishments that have been achieved over the past six years.

Expeditionary environments create special challenges for perception systems in autonomous ground vehicles. To address these challenges, a perception pipeline has been developed that fuses data from multiple sensors (color, thermal, LIDAR) with different sensing modalities and spatial resolutions. The paper begins with in-depth discussion of the multi-sensor calibration procedure. It then follows the flow of data through the perception pipeline, detailing the process by which the sensor data is combined in the world model representation. Topics of interest include stereo filtering, stereo and LIDAR ground segmentation, pixel classification, 3D occupancy grid aggregation, and costmap generation

A major contributor to path planning failure is costmap and world model noise and errors, which result when the characteristics of the current environment deviate from the data used to train the sensor perception system. Our adaptive traversability capability improves costmap estimation from purely passive sensing using an online learning strategy that uses successful traversal by the vehicle as a training signal to improve estimation over time. This adaptation mitigates the negative effects of noise and errors in generation of the world model and costmap and decreases dependency and sensitivity to fixed or user-set parameter values. Our system learns the association between traversability and a diverse set of descriptors that characterize the world model and costmap. These descriptors include single voxel-based quantities such as color and temporal coherence and collections of voxels over small local areas such as ground smoothness. By measuring or learning the data characteristics in the fixed sensor perception system and resulting world model, principled estimation of the correctness of the world model and costmap can also be incorporated. We avoid duplicating sensor processing directly in favor of characterizing the derived world model data. This allows the system to directly mitigate the noise and errors that can negatively impact path planning. The system design is general and modular, allowing easy exploration of different learning approaches. Both linear and non-linear methods have been examined, with non-linear methods showing more stable and better performance. We show successful terrain traversability estimation in a diverse set of environments, including unimproved road and expeditionary.

By employing selective reasoning for wheel placement on uneven terrain, planning systems generate safer trajectories that reduce overall mission risk by preventing high-centering and excessive wheel jolt, and therefore expand an autonomous vehicle’s mission portfolio by enabling them to navigate previously untraversable terrain. We show successful wheel placement reasoning in a diverse set of environments, including unimproved roads and expeditionary scenarios. We quantify the performance improvement by measuring ride safety measures (vertical acceleration and underbody clearance) and the number of vehicle collisions in a boulder corridor test environment.

Sensor augmentation with a priori data effectively expands the sensing range and adds new sensing modalities. High resolution elevation maps are registered to live estimates of the ground height to provide robust GPS estimates and to avoid steep slopes which may be unobservable by the sensors. Live sensor data is fused into persistent maps that are continuously relaxed and updated with new information. The map provides a robust location estimate and can be fused with the current live costmap to fill in gaps beyond the sensing horizon, allowing the navigation system to effectively re-route the vehicle over large distances.

Mission Modeling, Planning, and Execution Module (M2PEM) is a user friendly graphical framework for mission design and execution. It extends a subset of the Business Process Modeling and Notation (BPMN) 2.0 for robotic applications. Hierarchical abstractions fundamental to BPMN allow the mission to be naturally decomposed into interdependent parallel sequences of BPMN elements. M2PEM adapts these elements in a role based framework which uses collaborative control modalities as an atomic building block. Designed missions are able to consider situational data, external stimuli, and direct user interaction. Missions are directly executable using a resource manager and a ROS-based execution engine.

A novel cooperative path planning framework is presented for maintaining formations along a desired, unknown route where spatial and temporal objectives must be considered. It uses a reference frame based on longitudinal spacing along the route and lateral orthogonal offsets to plan for route clearance and traversal around previously unknown obstacles. By defining desired positions in terms of offsets from the route, the spatial and temporal components can be decoupled. The spatial components are planned using a two-step planner for fast real-time planning. The spatially defined paths are passed to a speed adaptation algorithm for temporal considerations including inter-vehicle collision avoidance. The approach balances real-time obstacle avoidance, spatial structure of the formation, and inter-vehicle safety considerations.

Interfacing with and supervising a team of heterogeneous, modular robotic systems across various, dynamic mission sets poses significant demands on the design of an operator control unit (OCU). This paper presents an architecture and OCU design which focuses on three key elements addressing this problem: 1. A well-defined system architecture for external interfacing, 2. Dynamic device discovery to advertise current capabilities, and 3. A flexible front-end to convey situational awareness. The system architecture and external interface is influenced by work done in 4D/RCS (Four Dimensional Real-time Control System) to provide a hierarchical decomposition of the system autonomy. The platform independent, web-based OCU is capable of interacting with any level of the system-of-systems from low-level teleoperation of a single vehicle to section-level concerns such as commanding coordinated tactical behaviors across multiple vehicles. The architecture design facilitates dynamic discovery of the team of vehicles and their respective payloads. The OCU subscribes to pertinent data from these advertised systems, providing multiple human supervisors the ability to interact with and receive concurrent visual representation of them and their status. System failures, payload detections, and user decision prompts are displayed in a prominent yet unobtrusive manner. The OCU was used to control a heterogeneous team of four vehicles performing coordinated route clearance operations in a set of four demonstration vignettes. Results of a technical exercise using this system are described herein. Keyword

Regimented experimentation of autonomous unmanned vehicles is an important enabler for performance assessment, issue identification, and root-cause analysis. As autonomous capabilities improve, these are increasingly problems of needles and haystacks; large amounts of test data are often required to fully identify and analyze erratic events. Acquisition and analysis of this test data can be arduous, error prone, and inefficient.

Driven by a philosophy that test methods should be repeatable, reproducible, and as automated as possible, we have developed an experimentation regime for expeditionary autonomous unmanned vehicles. This regime represents a method of decomposing functionality, building up evidentiary information, and analyzing component results on overall system performance despite the challenge of testing to sufficient statistical significance. We have also developed a set of methods and tools that automate data collection and analysis to facilitate experimentation. This paper describes this test and experimentation philosophy, regime, and associated tools developed under the Office of Naval Research, Code 30 Ground Vehicle Autonomy program.

The development of autonomous systems presents many challenges, including safety, repeatability of results, and the high cost of system testing. These challenges are compounded because of the nature of autonomous systems. The systems are composed of numerous interconnected components encompassing a broad spectrum of disciplines, including mechanical engineering, electrical engineering, computer vision, computer science, networking, and communications. By necessity, each component is often under simultaneous development with one or more other components. Because of the close coupling of components, changes in one component can have wideranging systemic effects. In this paper we describe a set of processes and tools that allow a distributed team to coordinate parallel development of software for autonomous systems without introducing system instability or performance regression. These include processes and tools for sharing software configurations, deploying software to autonomous system computers that lack internet access, building software in parallel across multiple machines, verifying software configurations, and automatically ensuring a specified level of quality exists using static analysis, unit tests, and regression tests. We build on existing best practices to ensure a stable baseline software configuration necessary for quantitatively assessing software changes without the need for impractically time-consuming data collection and testing. We have incorporated best practices from software development and robotics communities as well as from military requirements for fielding systems. These practices and tools give an increased confidence that the results from each test case capture the desired data due to proper software configuration, and that the configuration can be merged successfully into the baseline system if warranted by results. The processes described allow for parallel development paths to proceed simultaneously while minimizing the occurrence of unforeseen and difficult-to-resolve conflicts.

This paper describes new autonomy technology that enabled a team of unmanned surface vehicles (USVs) to execute cooperative behaviors in the USV Swarm II harbor patrol demonstration and provides a description of autonomy performance in the event. The new developments extend the NASA Jet Propulsion Laboratory’s CARACaS (Control Architecture for Robotic Agent Command and Sensing) autonomy architecture, which provides foundational software infrastructure, core executive functions, and several default robotic technology modules. In Swarm II, CARACaS demonstrated higher levels of autonomy and more complex cooperation than previous on-water exercises, using full-sized vehicles and real-world sensing and communication. The core autonomous behaviors to support the harbor patrol scenario included Patrol, Track, Inspect, and Trail, providing the capability of finding all vessels entering the patrol area, keeping track of them, inspecting them to infer intent, and trailing suspect vessels. Significantly, CARACaS assumed responsibility for not only executing tasks safely and efficiently but also recognizing what tasks needed to be accomplished, given the current state of the world. Since the heterogeneous USV teams shared world model that evolved, such as due to (dis)appearance of vessels in the area or a change in health or availability of a USV, CARACaS replanned to generate and reallocate the new task list. Thus, human intervention was never required in the loop to task USVs during mission execution, though a supervisory role was supported in the autonomy system for mission monitoring and exception handling. Finally, CARACaS also ensured the USVs avoided hazards and obeyed the applicable rules of the road, using its local motion planning modules.

The Office of the Under Secretary of Defense (OUSD) for Acquisition, Technology, and Logistics (ATandL) Joint Ground Robotics Enterprise (JGRE) and the Office of the Assistant Secretary of Defense for Research and Engineering (OASD RandE) support a new joint initiative to establish an unmanned systems communications framework called Joint Communications Architecture for Unmanned Systems (JCAUS). It is an architecture for a Family of Communications Systems (FoCS) that enforces industry standards, encourages rapid technology insertion, and fosters industry innovation and competition. This architecture adopts the Modular Open Systems Architecture (MOSA) principals aiming to reduce total ownership cost, accelerate technology transformation, and improve interoperability among the Department of Defense (DoD) Unmanned aerial/ground Systems (UxS) programs. This paper provides the background and technology highlights of the JCAUS program.

The Space and Naval Warfare (SPAWAR) Systems Center (SSC) Pacific is developing a prototype cybersecure network radio for the Man-Transportable Robotic System (MTRS) MK2 Mod 0 explosive-ordnance-disposal (EOD) robot that will also be compatible with the Advanced EOD Robotic System (AEODRS) Increment-2 robot. Both programs of record are managed by the Naval Surface Warfare Center (NSWC) Indian Head EOD Technology Division (IHEODTD).

This paper describes SSC Pacific’s solution, the MTRS/AEODRS Radio for Cybersecure Operation with Network Integration (MARCONI) system. MARCONI is a modular networking radio developed based on the Joint Communication Architecture for Unmanned Systems (JCAUS), which is being defined by the Joint Ground Robotics Enterprise (JGRE). The system architecture includes a Payload Host Module, which provides translations and interfaces to the other modules, a Communication Module based on frequency-shifted and amplified 802.11n wireless technology with ad hoc networking, and a Cybersecurity Module, which uses the latest technologies—including Tactical Key Management—to meet the unique requirements for unmanned systems.

Unmanned Systems (UxS) present unique challenges to Cybersecurity developers due to: (1) The need to secure these systems and protect classified information in the unmanned operational environment, where the UxS are often outside the visual range and physical control of the operator. (2) They require very low Size, Weight, and Power (SWaP) consumption constraints imposed by the smaller UxS. (3) They also impose long and costly approval processes for employing Cryptographic Products, Key Management Architectures, and other High Assurance security solutions in the military operational environment. Additionally, modularizing communications architectures to take advantage of rapid technological and SWaP advancements often results in multiple classification domains, and the need to move plaintext control information between these domains, which results in a Cross Domain Solutions (CDS), requirement.

This paper details work supporting High Assurance Cybersecurity engineering for UxS that includes: (1) Investigating current approved cryptographic and CDS tactical products to assess their suitability for UxS operations. Although some current products can meet the larger UxS SWaP requirements, most approved products do not meet SWaP requirements for the smaller UxS. (2) Exploring potential new technologies that have not yet reached the operational stage but show promise for meeting the smaller UxS SWaP requirements. (3) Evaluating potential solutions that will meet the three challenges of deploying High Assurance Cybersecurity solutions onboard UxS in military operations.

To lighten the load of dismounted infantry, research is being conducted into the use of “robotic mules” that can help carry equipment and supplies for small units as they conduct their operations. While autonomy and perception features of robotic systems are steadily improving, the way in which people interact with them is still fairly rudimentary. The operator control units (OCUs) for these robotic mules are typically hand-held gaming controllers for tele-operation, or worn ruggedized portable computers or tablets with point-and-click interfaces. These control devices add more weight to the operator, and often require them to hold the controllers in their hands and to spend considerable time looking down at the OCU screen to enter commands or to understand what the robot is doing. Furthermore, these interfaces often require specialized training to understand how to operate the OCUs. This paper describes research aimed at reducing the physical, cognitive, and training burdens that robotic systems place on operators above and beyond the their regular jobs as warfighters. We first present an analysis of relevant infantry communication to identify interaction requirements, and an analysis of technologies that might be used to support these interactions. We then describe a prototype heads-up, hands-free system for controlling robotic mules using a lightweight, worn interaction device that facilitates natural twoway interaction (including speech and gesture input) between the robotic mule and the user. We describe the challenges in building this system and some formative evaluations of the technology

Remotely controlled or tele-operated robots are playing an increasing role in military, law enforcement and industrial operations. In order for a remote operator to rapidly and efficiently control an unmanned vehicle (UV), the operator must be able to view the robot surroundings as if they were actually present at the location of the robot. The current limiting factor to providing this level of presence is the visual display utilized by the operator, which for portability, size and weight is optimally a head-worn display (HWD). When in addition local situational awareness of the operator must be maximized, the HWD should be of the optical see-through (OST) type. The ideal requirements for the OST-HWD for UV-control include: a large binocular display field-of-view, methods to prevent development of cyber sickness and appropriate brightness for use in all relevant lighting conditions. Limitations of current OST-HWDs are reviewed, and a new type of OST-HWD is described, which offers significant advantages for remote control of UVs. An initial prototype has been constructed, and its performance is quantified. The prototype provides a large display FOV, large eye box, and high brightness. It uses low-cost COTS components. When the operator of the unmanned vehicle must use their hands for other purposes, a hands-free control interface for both the display and robot is highly desirable. An implementation of a hands-free control interface based on eye tracking is described.

This paper overviews the development and operator testing of a shared autonomy system for small unmanned ground vehicles operating in indoor environments. The project focused on creating driving assistance technologies to reduce the burden of performing low-level tasks when operating in cluttered or difficult areas by sharing control between the operator and the autonomous software. The system also provides a safety layer to prevent the robot from becoming disabled due to operator error or environmental hazards. Examples of developed behaviours include obstacle proximity warning, centering the vehicle through narrow doorways, wall following during long traversals, tip-over indicator, stair climbing aid, and retreat from communications loss. The hardware and software were integrated on a QinetiQ Talon IV robot and tested by military operators in a relevant environment.

This paper presents a new obstacle avoidance method that provides unobtrusive assistance to tele-operation of unmanned ground vehicles. Different from existing obstacle avoidance methods, the present method can determine whether the driving commands from an operator are safe in the presence of obstacles and can automatically adjust unsafe commands to help the operator avoid proximate obstacles. The command adjustment is done in an unobtrusive manner and conforms to the dynamic and kinematic constraints of the vehicle in order to minimize its interference to the operator. Due to its assistive and unobtrusive nature, this method can quietly share some control authority with an operator in tele-operating a vehicle, and hence it has the potential to make tele-operation of ground vehicles in challenging environments significantly easier and safer. The effectiveness of this method is demonstrated in extensive experiments in cluttered environments using military-grade tracked robots.

The Autonomous Ground Resupply (AGR) Program is a U.S. Army Tank Automotive Research Development and Engineering Center (TARDEC) effort to reduce the number of soldiers required for ground resupply. One objective of this program involves providing vehicles with the capability to operate unmanned in a variety of circumstances. Prior to fielding, this system is undergoing an assessment to identify possible vulnerabilities. We use this system to illustrate a general approach to the assessment of the autonomous decision making, describing a process which could be used to identify vulnerabilities in the autonomy of military systems without describing any actual vulnerabilities discovered in the AGR system.

The IGVC is a college level autonomous unmanned ground vehicle (UGV) competition that encompasses a wide variety of engineering professions – mechanical, electrical, computer engineering and computer science. It requires engineering students from these varied professions to collaborate in order to develop a truly integrated engineering product, a fully autonomous UGV. Students must overcome a large variety of engineering technical challenges in control theory, power requirements/distribution, cognition, machine vision, vehicle electronics, sensors, systems integration, vehicle steering, fault tolerance/redundancy, noise filtering, PCB design/analysis/selection, vehicle engineering analysis, design, fabrication, field testing, lane-following, avoiding obstacles, vehicle simulation/virtual evaluation, GPS/waypoint navigation, safety design, etc.

A robust onboard full angular orientation sensor solution, based on a polarization scanning reference source and polarized geometrical cavity orientation sensors is presented. The sensor system provides a new non-GPS and non-inertial approach to angle measurements, with several key advantages over other methods, which include traditional phased-array antenna systems and the like. In this novel approach, the angular orientation information is coded into a time dependent pattern, which is insensitive to noise, while making the angle measurement independent of distance from the referencing source. The manuscript presents data from a field deployable system.

In this paper we propose a trust algorithm, dubbed NeuroTrust, based on a multi-layered neural network. Previous work introduced trust as a performance estimation algorithm between team members in multi-agent systems, to allow for behavior optimization of the team. The trust model was developed based on an Acceptance Observation History (AOH) and confirmation and tolerance parameters to control trust growth and decay. Further work proposed certain improvements, in an autonomous vehicles convoy scenario, by considering agent diversity and a non-linear relationship between trust and vehicle control. In this work we show a further optimization using a deep recurrent neural network. This multi-layered neural network delivers trust as a probability function estimation with AOH as a sliding window batch input. The neural network is pre-trained using supervised learning, to emulate the previous trust model, as baseline. This pre-trained model is then exposed to future optimization using on-line reinforcement learning. The proposed trust model could be adaptable to a variety of systems, external conditions, and agent diversity. One application example where such a biologically-inspired trust model is suitable would be for soldier-machine teaming. Furthermore, particularly in the autonomous convoy scenario, we can account for the trust-control relationship nonlinearity in the trust domain, thus simplifying the control algorithm.

Military applications require networked multi-UAV system to perform practically, optimally and reliably under changing mission requirements. Lacking the effective control and communication algorithms is impeding the development of multi-UAV systems significantly. In this paper, distributed optimal flocking control and network co-design problem has been investigated for networked multi-UAV system in presence of uncertain harsh environment and unknown dynamics. First, the mathematical interaction between network imperfections and practical wireless network channel performance has been investigate. Then, a novel co-model has been developed for networked multi-UAV combining effects from physical system and network channel model effectively. Then, adopting neuro dynamics programming (NDP) technique and actor-critic-identifier (ACI) design architecture, a novel online finite horizon optimal flocking control and network co-design has been proposed. The developed algorithm cannot merely obtain the distributed optimal co-design within finite time, but also relax the stringent requirement about physical UAV system and network dynamics. In addition, developed novel co-design can satisfy the practical constraints, e.g. transmit power constraint etc. The Lyapunov stability analysis is used to validate the effectiveness of developed scheme. With the proper NN weight update law, proposed co-design can ensure all closed-loop signals and NN weights are uniformly ultimately bounded (UUB). Furthermore, simulation results have been provided to demonstrate the effectiveness of the developed scheme.

Due to the highly uncertain and dynamic nature of military conflict and planetary exploration missions, enabling aerial and terrestrial unmanned autonomous systems (UAS) to gracefully adapt to mission and environmental changes is a very challenging task. In particular, the United States Army, Air Force, Navy, and NASA have recently shown interest in the task of load transportation by means of UAS, which rely heavily on the knowledge of both the UAS model and the load dynamics to function. Most of the currently available autopilot systems for UAS were built without suspended load transportation capabilities and are thus not appropriate, for example, to assist soldiers or planetary explorers in the tasks of carrying and deploying supplies, transporting injured people, or warfare. This research provides knowledge to the problem of autonomous suspended load transportation, attending national agencies expectations that UAS will perform in a reliable manner even in the challenging situation when loads of uncertain characteristics are transported and deployed, which heavily modify the UAS dynamic during the execution of the task. This work presents a novel model-free adaptive wavenet PID (AWPID)-based controller for enabling aerial UAS to transport cable suspended loads of unknown characteristics. In order to accomplish this goal, a control design is presented which enables the UAS to perform a trajectory tracking task, based solely on the knowledge of the UAS position. The methodology proposes a novel structure, which identifies inverse error dynamics using a radial basis neural network with daughter Mexican hat wavelets activation function. A real-time load transportation mission consisting of a multi-rotorcraft UAS carrying a cable suspended load of unknown characteristics validates the effectiveness of the trajectory tracking control strategy, showing smooth control signals even when the mathematical model of the aerial UAS and load dynamics are not known.

In time critical power up applications, such as pre-launch preparation of smart munitions, both power and guidance information needs to be transferred rapidly from the host to the munitions. Tethered solutions are not practical and the existing magnetic inductive charging methods are slow and have limited data transfer rate capability. A wireless solution based on the use of high power laser diodes is presented. Custom dual-junction photo-voltaic cells achieve power conversion efficiencies exceeding 50% at a single wavelength of operation. Energy transfer times of 3.5 s have been achieved for energy levels of 90 J. Guidance and other control data is uploaded to on-board memory devices at a rate of 500 kb/s, through the use of an additional photodiode, which can receive a small fraction of the modulated power beam. A removable collar provides an alignment free charging/data solution enabling rapid deployment of multiple munitions.

The ability to upload a space or unmanned aircraft mission as a self-contained software application can enable multiple users to share an aerial, orbital sensing, or other platform. However, incorporating this mechanism presents the risk of uploading malicious software to the device and damaging or impairing the system. The App-On-Demand capability provides an answer to this. It allows applications to be uploaded to the craft, run, and have their data be returned to the ground, safely and securely. This paper provides an overview of the design and development of the App-On-Demand functionality for the OpenOrbiter I spacecraft.

Most spacecraft will encounter an error at some point in their flight and that error needs to be addressed for the spacecraft to function as it was originally intended. The goal of this paper is to create a possible system capable of detecting errors and fixing them. To do this, the system must first detect the errors. On-board software will detect the logical faults stemming from errors in the electrical components, and imaging will be able to detect the physical faults. The system will have a memory of many possible errors and how to fix them so that when it detects an error, it can repair the problem using the knowledge pre-programmed into it. During the repair, the system will have sensors to find and repair the fault.

UAVs path planning is usually based on availability of GPS signals for real time positioning. However, there many situations where the GPS signal might not be available either due to interference, or lack of enough number of available satellites at the desired area. In urban environment, the GPS signals may not be available due to high rise building. However, such environment is usually dense with cellular signals from largely distributed cellular tower all over the area. The cellular signals from different technologies such as CDMA and LTE are already available in addition to long time existing GSM technology. In this work, we propose to use the cellular signals from different towers and sectors to provide localization information to the UAV in addition to environment canopy database where heights of building, trees, etc are used obstacle database in path planning algorithm. The work is to show how to build three dimensional radio frequency maps either for one or different modes of technologies and how these RF maps can be used by UAV to estimate its position in 3D space and used by the guidance, navigation and control system of the UAV to position itself and determine trajectory selection in three dimension space of an urban environment.

As more vehicles and devices are becoming equipped with cameras, computer vision technologies can significantly enhance functionalities. Among various computer vision applications, alphabet recognition has been an important application. Alphabet recognition is a basic operation in text recognition of natural images, which can be used to obtain useful information from visual information. It can be used for vehicles or autonomous moving machines such as drones or robots to understand surrounding environments. A main difficulty is that alphabets can take various sizes and orientations. Thus, it is desirable to develop recognition algorithms that are size and rotation invariant. In this paper, we propose alphabet recognition algorithms based on the recently proposed angle-distance map, which is robust against size and rotation variations. Using the two features (distance and angle), the angle-distance map is generated, and a rotation invariant matching algorithm is developed. Since the distance is normalized, the map is size-invariant. Experimental results showed promising results.

The necessity and occurrence of Unmanned Aerial Vehicles on the battlefield in near future is increasing day-by-day and that leads to the problem of sharing air space with manned air vehicles. However, avoiding collisions and/or deploying countermeasures during threat detection are a very crucial issue in most of the unmanned aerial vehicles [1] or similar. There is a need for a Sense, Avoid or React [2] system to track objects of potential collision risk or determine any action to avoid or mitigate a collision and, react with countermeasures after detection of hazardous situations i.e., during midair attack, collisions, flight path obstacles or dense clouds. The author of this paper present an algorithm for decision making system based on countermeasures during collision avoidance scenario or threat detection. A general framework to deal with non-linear dynamic systems and will developed consisting of a system of various collision and risks scenarios with moving and stationary threats that is based on straight future projection. The solution will include an algorithm that captures the path prediction of the threat. The proposed optimization problem resolution will aim to maintain minimum separation between two vehicles or threats and applying necessary countermeasures if the collision becomes unavoidable. The multi-sensor fusion system, will generate the object status signal, merges vehicle status, will release an assessment of collision in the form of warning level. A fuzzy controller for countermeasure of friendly maneuver will be presented to generate active and passive measure signals based on the response from assessment signal and vehicle sensor signals. The design implementations and simulations using FPGA will be included.