This PDF file contains the front matter associated with SPIE Proceedings Volume 11021, including the Title Page, Copyright information, Table of Contents, Author and Conference Committee lists.

Oak Ridge National Laboratory presents a new UAS-mounted multi-modal imaging payload containing five sensors. We have integrated several distinct commercially available sensors onto a large Class-1 autonomous quadcopter aircraft: a LIDAR (Light Detection and Ranging) scanner, a hyperspectral pushbroom sensor, a multispectral camera, a longwave infrared thermal camera, and an RGB camera. The system integrates our proprietary Multi-modal Autonomous Vehicle Network (MAVNet) and communication system, allowing autonomous control via multiple communication networks. Using one common Global Navigation Satellite System (GNSS) and inertial navigation system (INS/GPS), imagery from all sensors are accurately and precisely geolocated and co-registered.

Unmanned aerial vehicles (UAVs) or drones are rapidly gaining popularity in the field of remote sensing for capturing images with ultra-high spatial resolution while flying at lower altitudes. Development of highlyefficient miniaturized sensors and the use of geospatial image processing techniques have been immensely helpful in growing this technology as the most sought-after and sophisticated remote sensing technique at a relatively lower cost. Parking lot occupancy detection, one of the current drone based application area, can be used for effective management of parking spaces, to reduce queues, minimize the time required to find an area, and to issue tickets in cases of parking violations. Visual information based parking lot monitoring techniques are mostly tailored for specific applications, and they lack generalization. In this paper, a UAV-assisted quick and efficient monitoring solution is proposed for real-time parking occupancy and car number plate detection. In this approach, a drone-mounted camera has been used to capture images of the parking lot under consideration. Initially, the parking lot is being mapped using drone-coordinates, and then a dynamic programming algorithm is used to determine the shortest route to cover multiple parking lots in minimum time to capture maximum number pictures solved by the Traveling Salesman Problem (TSP). For capturing images, an optimum location is used, such that the drone can cover the maximum area of a parking space in minimum time. Depending on the parking area under investigation, paths altitudes and gimbal angles of the drone is changed dynamically while capturing images. Then a deep neural network based parking lot occupancy monitoring system is used to determine the number of occupied and vacant spots in a parking lot. The drone captured images of each parking spot are then tested with a pre-trained model based on car and non-car images. Then the automatic license plate recognition (ALPR) algorithm is used for parking rule enforcement. Finally, experimental results are verified using a web based application that is connected with a cloud database.

Recent military research has focused on how militaries can effectively employ technology in mega-cities deemed Contested Urban Environments (CUE). This paper looks at aspects of the CUE environment and discusses how autonomous systems may be used to gain a tactical advantage. Recent experiments at The Technical Cooperation Program (TTCP) CUE 18 experiment, held in Montreal, looked at the use of autonomous systems, among other technologies, in roles such as surveillance, mapping, sensor deployment, etc. A key focus of the work was to improve the situational awareness of the soldiers at the tactical edge. Employment and delivery of unmanned assets at the tactical edge demands robustness to communication and GPS loss, ease of use, and full-scale integration with larger command and control structures to ensure maximum effectiveness of the assets. Several systems were deployed at the CUE Montreal trial to test these environmental stressors. For instance, UAVs carrying vision and lidar-based sensors were used to produce detail 3D maps of the environment without relying on GPS information. A novel concept for deploying Unattended Ground Sensor (UGS) from a UAV was developed and used to deploy the sensors to strategic locations. Data from the unmanned vehicles and the UGS were shared through the use of the Open Standards for Unattended Sensors (OSUS) and accessible to both the Canadian and allied Common Operating Picture (COP) being tested at the trial. The remainder of this paper discusses the results of the CUE Montreal experimentation and uses the results to suggest future directions of autonomous research in CUE environments.

Mobility and terrain are two sides of the same coin. We cannot speak to our mobility unless we describe the terrain’s ability to thwart our maneuver. Game theory describes the interactions of rational players who behave strategically. In previous work¹ we described the interactions between a mobility player, who is trying to maximize the chances that he makes it from point A to point B with one chance to refuel, and a terrain player who is trying to minimize that probability by placing an obstacle somewhere along the path from A to B. This relates to the literature of games of incomplete information, and can be thought of as a more realistic model of this interaction. In this paper, we generalize the game of timing studied in the previous paper to include the possibility that both players have imperfect ability to detect his adversary.

Unmanned Aerial Vehicles (UAVs) are very popular and increasingly used in different applications. For many applications, it can be very interesting to achieve UAVs collaboration. In this work, we propose the use of vision-based collaboration between UAVs. The proposed approach uses images captured by a UAV and deep learning to detect and follow another UAV. To detect the leader UAV, we developed an approach based on the deep YOLO algorithm. This approach was able to process videos at 30 fps and get high mAP for UAV detection. To follow the leader UAV, we developed a high-level control algorithm based on the use of the detected bounding box coordinates. The bounding box size and position help compute the command to send to the follower UAV. Tests were conducted in outdoor scenarios using quadcopter UAVs. The obtained results and the high mAP are promising and show the possibility of using this kind of vision-based deep learning approach for UAVs collaboration.

Over the past few years, UAVs have known and increase in popularity and are now widely used in many applications. Today, the use of multiple UAVs and UAV swarms are attracting more interest from the research community leading to the exploration of topics such as UAV cooperation, multi-drones autonomous navigation, etc. In this work, we are interested in UAVs tracking and pursuit. The goal here, is to use deep learning and the captured images from one of the UAVs to detect and track the second moving UAV. The proposed approach uses deep reinforcement learning for UAV pursuit. The input is the current frame cropped using the last target pose, and the output is a probabilistic distribution between a set of possible actions. The experimental results are promising and show that the proposed algorithm achieves high performances in challenging outdoor scenarios.

Autonomous bicycles offer numerous potentials for smart city applications thanks in part to their light weight, safe autonomy, being optionally manned, and last-mile delivery. This paper describes the design of a self-stabilizing autonomous bicycle with electric linear actuators. The high-speed linear actuator is mounted between the seat and the handlebar of the autonomous bicycle, which provides the bicycle with high peak power and energy efficiency. Physical tests are carried out to verify automatic steering and speed regulation capabilities of the autonomous bicycle.

Riderless bicycles, which belong to the class of narrow autonomous vehicles, offer numerous potentials to improve living conditions in the smart cities of the future. Various obstacles exist in achieving full autonomy for this class of autonomous vehicles. One of these significant challenges lie within the synthesis of automatic control algorithms that provide self-balancing and maneuvering capabilities for this class of autonomous vehicles. Indeed, the nonlinear, underactuated, and non-minimum phase dynamics of riderless bicycles offer rich challenges for automatic control of these autonomous vehicles. In this paper, we report on implementing linear parameter varying (LPV)-based controllers for balancing our constructed autonomous bicycle, which is equipped with linear electric actuators for automatic steering, in the upright position. Experimental results demonstrate the effectiveness of the proposed control strategy.

Miniature blimps will have numerous applications in future smart cities. This paper presents the design of an autonomous blimp that can be autonomously operated and controlled. In order to be able to operate over long periods of time, the blimp design employs a novel actuation mechanism with only one servomotor and two DC motors. Experiments are carried out to demonstrate the capabilities of the constructed autonomous blimp.

This paper presents results from an experiment performed at the Combat Capabilities Development Command, Army Research Laboratory, Autonomous Systems Division (ASD) on the precision of a 7-degree-of-freedom robotic manipulator used on the RoMan robotic platform. We quantified the imprecision in the arm end-effector final position after arm movements ranging over distances from 362 mm to 1300 mm. In theory, for open-loop grasping, one should be able to compute the final X-Y-Z position of the gripper using forward kinematics. In practice, uncertainty in the arm calibration induces uncertainty in the forward kinematics so that it is desirable to measure this imprecision after different arm calibrations. Forty-one runs were performed under different calibration regimes. Ground truth was provided by measuring arm motions with a Vicon motion capture system while the chassis of the platform remained stationary during the experiment. Using a digital protractor to align the arm joints to the ground plane for a “Level” type calibration, the average total offset of the gripper in 3D space was 19.6 mm with a maximum of about 30 mm. After a “Field” (i.e. Hand-Eye) calibration, which aligned fiducials on the joints, the average total offset came to 37.8 mm with a maximum of about 80 mm. Distance travelled by the arm was found to be uncorrelated with total offset. The experiment demonstrated that Total (X, Y, Z) Offset in the gripper final position is reduced significantly if the robot arm is first calibrated using a standard “Level” calibration. The “Field” calibration method results in a significant increase in Offset variation.

Under the Navy’s Flexible Cyber-Secure Radio (FlexCSR) program, the Naval Information Warfare Center Pacific and the Massachusetts Institute of Technology’s Lincoln Laboratory are jointly developing a unique cybersecurity solution for tactical unmanned systems (UxS): the FlexCSR Security/Cyber Module (SCM) End Cryptographic Unit (ECU). To deal with possible loss of unmanned systems that contain the device, the SCM ECU uses only publicly available Commercial National Security Algorithms and a Tactical Key Management system to generate and distribute onboard mission keys that are destroyed at mission completion or upon compromise. This also significantly reduces the logistic complexity traditionally involved with protection and loading of classified cryptographic keys. The SCM ECU is on track to be certified by the National Security Agency for protecting tactical data-in-transit up to Secret level. The FlexCSR SCM ECU is the first stand-alone cryptographic module that conforms to the United States Department of Defense (DoD) Joint Communications Architecture for Unmanned Systems, an initiative by the Office of the Secretary of Defense supporting the interoperability pillar of the DoD Unmanned Systems Integrated Roadmap. It is a credit cardsized enclosed unit that provides USB interfaces for plaintext and ciphertext, support for radio controls and management, and a software Application Programming Interface that together allow easy integration into tactical UxS communication systems. This paper gives an overview of the architecture, interfaces, usage, and development and approval schedule of the device.

Small Unmanned Aerial Systems (UASs) have great potential for many different applications [1- 5]. The small UASs are lightweight, man-portable, and capable of carrying payloads. For military applications, these systems provide valuable intelligence, surveillance, reconnaissance, and target acquisition (ISRTA) capabilities for units at the infantry battalion, company, and platoon levels. The power system is a key component for small UASs to perform extended and long-range missions. We have selected, examined, developed, and evaluated several cutting-edge power and energy technologies to power small UASs. Currently, the capabilities of a small UAS are limited by its power source. Small UASs are mainly powered by advanced batteries, which cannot sustain extended operations. Small engine generators are not a viable solution because they generate pollutants and can be noisy, which could be detected by the adversary. Solar cells are not efficient enough to be used as the primary power system and are limited by weather conditions. Polymer Electrolyte Membrane Fuel Cells (PEMFCs) still have the same technical constraint, the source of hydrogen, as they did many years ago. The objective of this work is to develop, demonstrate, and integrate a highly efficient, lightweight 350 W Solid Oxide Fuel Cell (SOFC) system for small UAS applications. The result of this developmental effort will be a power system to support increased mission duration, power, and reliability of the small UAS, resulting in improved situational awareness. Improved situational awareness capabilities will specifically benefit Department of Defense convoys, route clearance missions, base/defense patrols, and other reconnaissance objectives. The research and development efforts presented here not only apply to small UASs but can also help extend mission operations for unmanned ground vehicle systems and Soldier-portable power application.

Consumer drone technology is a rapidly evolving field and expected to revolutionize the world in wide aspects. Drones’ air traffic control is coming in near future with high priority requirement as part of national airspace system. Massive number of flying drones require frequency band that can support high throughput and very short latency. Current UHF frequency bands reach their capacity limit. In this work, we investigate propagation characteristics of the radio channel in super high frequency (SHF) band as the expected operating band for wireless controlled drones. The presented radio channel characteristics cover effect of frequency range, in terms of path loss and pattern propagation factor behavior with vertical and horizontal polarization.

Unmanned aerial vehicles, commonly known as drones, are poised to provide numerous beneficial services ranging from package delivery to surveillance. However, they also represent a threat when used for nefarious purposes. Problematic applications range from illegal activities such as smuggling to those that may be dangerous, such as flying over people at an event, to annoying and invasive activities, such as a neighbor spying. In cases where there are national security or safety implications or a need to stop an illegal activity, there is significant benefit in being able to disable, capture or even destroy drones. To this end, numerous technologies for interfering with drone operations and trying to capture drones have been proposed. However, one of the most globally effective approaches, which can be as useful on a handheld drone as on one with dozens of feet of wingspan, is attacking its control and communications systems: that is, a cyberattack. This paper discusses how drones can be attacked electronically and techniques which can be designed to confuse, gain control of and even damage or destroy drones. In particular, numerous types of common exploits are considered and their applicability to the drone attacking challenge is evaluated. The paper also discusses similarities and differences between the presented classes of anti-drone techniques and presents a decision matrix for choosing between them. It also discusses how drones can be hardened against cyber-attacks and the implications of this turning into a persistent battle of one-upmanship between drone attackers and defenders, over time.

In this paper, we present results obtained through simulation of connected-autonomous semi-trucks that are operating in a leader-follower configuration. Autonomy is enabled in this configuration with a very lean sensor package on each truck, precision global positioning system (GPS), radar-based automated cruise control system (ACC), and dedicated short-range vehicle-to-vehicle communication system (DSRC). Our simulation includes modeling the operating environment, namely, the high-speed test track at the American Center for Mobility (ACM); the sensors, namely, GPS, ACC, and DSRC; and vehicle dynamics of semi-trucks. Simulation results in this paper are focused on measuring the safety margin of the follower semi-truck under different environmental conditions. We studied adverse weather and measured the decrease in safety margins with the increase in precipitation.

Military planners envision a future of unmanned vehicle swarms that can self-organize to provide superior intelligence and overwhelming effects over a widely dispersed battlefield greatly multiplying the effectiveness of the manned forces. Deploying swarms requires new tactics to take advantage of this capability. Working with military experts we developed a suite of swarm tactics for unmanned air and ground vehicles supporting a full Company with a mission to secure an objective in an urban area. The air vehicles create and maintain perimeter security around the objective. They map the area and maintain tracks on all vehicles and people in the perimeter. They also provide persistent, stealthy communication relay support to all the ground forces. The ground vehicles surveil the key intersections, provide scout and rear security to the squads and scout out alternative routes through the city. The unmanned vehicles execute decoy operations and their behaviors are designed to mask the actual task they are performing through seemingly random movement and regular swapping of tasks. These tactics were implemented in software and evaluated in a 3D model of an urban area taken from a city in the Midwest and implemented in Unity. This paper describes the tactics, algorithms, and experimental setup and reports the results.

We review efforts by NAVAIR and NASA Jet Propulsion Laboratory to automate the operation of the largest fleet of autonomous maritime vehicles. The vehicles are intended for large-scale demonstrations of US Navy systems and tactics to counter asymmetric naval threats. This review covers a distributed architecture for human-in-the-loop control of several autonomous high speed boats. Hazard avoidance and formation control was verified using real-world vehicles.

Advances in unmanned systems enable smaller, less expensive platforms that can be deployed in high numbers or swarms providing superior intelligence and overwhelming effects over a widely dispersed battlefield greatly multiplying their effectiveness. But swarming system remain a laboratory curiosity with only a few live demonstrations of limited scope. Swarms are complex and their behavior is difficult to predict requiring skilled engineers and hand-tuning for each mission. Based on 20 years of experience designing swarms for military missions, the Design of Self-Organizing Adaptive Robotic Swarms (DSOARS) is an engineering environment which addresses the three core challenges of swarm design: (1) decomposing mission tasks into the behaviors of the swarm entities, (2) configuring the size of the swarm to a specific mission, and (3) verifying that the resulting swarm behavior consistently achieves the mission goals with a high level of confidence. DSOARS addresses these challenges through two primary innovations: (1) a means to create verified swarm design patterns that decompose high level mission tasks into individual behaviors and (2) a constructive test environment that simultaneously optimizes and characterizes the swarm performance against a range of possible mission conditions. Users with no swarm expertise can specify the requirements and constraints of their mission and DSOARS will configure a swarm that can meet those objectives with performance guarantees. This paper describes the approach and reports experimental results building and configuring a suite of swarm tactics for an urban mission.

In previous work, a multi-layered neural network trust model, dubbed NeuroTrust, was introduced. This trust model was also implemented in an autonomous vehicles convoy simulation, in which speed and gap distance depended on trust. It has been shown that, in time, through on-line reinforcement learning, this trust model produces better results for significant performance metrics in the respective autonomous vehicle convoy when compared to a baseline trust algorithm. In this paper, the NeuroTrust model is expanded to leverage the experience of multiple decision-making agents. A trust aggregation method is proposed for NeuroTrust and is simulated for multiple autonomous vehicle convoy scenarios. It is shown that the NeuroTrust model tends to optimize faster by leveraging each agent’s experience.

Modern intelligent systems are expected to be able to learn from experience, making decisions on the basis of the available information and proceeding step by step to a desired goal. An important specification of such adaptive decision making method is the amount of time to accomplish a decision. In this paper, we propose a random walk model for such decision making method. The model involves random processes which have independent stationary increments. The decision times are formulated as first passage times dependent on the parameters of decision rules. Asymptotic and nonasymptotic results are developed for the analysis of first passage times.

A persistent concern of control engineering is the performance of systems in the presence of uncertainty. In this paper, we consider uncertainties affecting systems as stochastic processes of independent stationary increments. We show that in many situations the performance of an uncertain system can be measured by a function of a parametric stopping time and associated values of stochastic processes. Under some mild regularity conditions, we demonstrate that the performance measure is governed by stochastic functional central limit theorems as the parameters of the stopping time tend to certain values. Such results can be applied to the analysis and design of control systems affected by uncertainties.

A critical issue affecting the success of decision making is the underlying uncertainty. In this paper, we consider decision making problems involving uncertainties characterized by stochastic processes of independent stationary increments. The cost function of decision making is expressed as a function of the decision time and associated values of stochastic processes. The decision time is a stopping time dependent on the parameters of decision rules. We investigate the asymptotic behavior of the cost function as the parameters of decision rules tend to certain values. We demonstrate that the cost function follows stochastic functional limit theorems as the parameters of the decision rules tend to certain values.

Modern intelligent systems highly depends on their capabilities to learn from experience and take control actions in uncertain environments. In this paper, we propose a random walk approach for analyzing the performance of learning and control of intelligent systems. We show that in many situations, the learning and control problem can be formulated as a random walk in a hyperspace with stopping boundary defined by parameters of learning and control policies. We show that the performance of the intelligent systems can be measured by a function of the stopping time and associated values of stochastic processes. Under some mild regularity conditions, we demonstrate that the performance measure follows stochastic functional laws of the iterated logarithm as the parameters of learning and control policies tend to certain values.