The DARPA Robotics Challenge (DRC) program conducted a series of prize‐based competition events to develop and demonstrate technology for disaster response. This article provides the official and definitive account of DRC Finals as the culmination of the DRC program. The article details the eight tasks (Drive, Egress, Door, Valve, Wall, Surprise [Plug and Switch], Rubble [Obstacle or Debris], and Stairs) constituting the Challenge, and describes how the competition encouraged supervised autonomous operation by intentionally degrading the communications channel between the remote human operators. The article presents the results of the DRC Finals and places those results in perspective by identifying both strengths and weaknesses of robot performance exhibited at the competition.

This paper presents the software framework established to facilitate cloud-hosted robot simulation. The framework addresses the challenges associated with conducting a task-oriented and real-time robot competition, the Defense Advanced Research Projects Agency (DARPA) Virtual Robotics Challenge (VRC), designed to mimic reality. The core of the framework is the Gazebo simulator, a platform to simulate robots, objects, and environments, as well as the enhancements made for the VRC to maintain a high fidelity simulation using a high degree of freedom and multisensor robot. The other major component used is the CloudSim tool, designed to enhance the automation of robotics simulation using existing cloud technologies. The results from the VRC and a discussion are also detailed in this work. Note to Practitioners - Advances in robot simulation, cloud hosted infrastructure, and web technology have made it possible to accurately and efficiently simulate complex robots and environments on remote servers while providing realistic data streams for human-in-the-loop robot control. This paper presents the software and hardware frameworks established to facilitate cloud-hosted robot simulation, and addresses the challenges associated with conducting a task-oriented robot competition designed to mimic reality. The competition that spurred this innovation was the VRC, a precursor to the DARPA Robotics Challenge, in which teams from around the world utilized custom human-robot interfaces and control code to solve disaster response-related tasks in simulation. Winners of the VRC received both funding and access to Atlas, a humanoid robot developed by Boston Dynamics. The Gazebo simulator, an open source and high fidelity robot simulator, was improved upon to met the needs of the VRC competition. Additionally, CloudSim was created to act as an interface between users and the cloud-hosted simulations. As a result of this work, we have achieved automated deployment of cloud resources for robotic simulations, near real-time simulation performance, and simulation accuracy that closely mimics real hardware. These tools have been released under open source licenses and are freely available, and can be used to help reduce robot and algorithm design and development time, and increase robot software robustness.

The Defense Advanced Research Projects Agency (DARPA) has funded innovative scientific research and technology developments in the field of brain-computer interfaces (BCI) since the 1970s. This review highlights some of DARPA's major advances in the field of BCI, particularly those made in recent years. Two broad categories of DARPA programs are presented with respect to the ultimate goals of supporting the nation's warfighters: (1) BCI efforts aimed at restoring neural and/or behavioral function, and (2) BCI efforts aimed at improving human training and performance. The programs discussed are synergistic and complementary to one another, and, moreover, promote interdisciplinary collaborations among researchers, engineers, and clinicians. Finally, this review includes a summary of some of the remaining challenges for the field of BCI, as well as the goals of new DARPA efforts in this domain. (C) 2014 The Authors. Published by Elsevier B.V.

The U.S. Defense Advanced Research Projects Agency's (DARPA) Neovision2 program aims to develop artificial vision systems based on the design principles employed by mammalian vision systems. Three such algorithms are briefly described in this paper. These neuromorphic-vision systems' performance in detecting objects in video was measured using a set of annotated clips. This paper describes the results of these evaluations including the data domains, metrics, methodologies, performance over a range of operating points and a comparison with computer vision based baseline algorithms.

Abstract : The purpose of this project was to develop methods and software to determine whether a given Small Unmanned Ground Vehicle (SUGV) can traverse a given terrain, when both the SUGV and the terrain are not known exactly. A simulation model of a real-world SUGV (iRobot PackBot) was developed in the ADAMS environment and used to simulate traversal of a variable-height step obstacle. For this project, a user subroutine was successfully integrated into the ADAMS model to predict deformable track-terrain interaction. A parameterizable UGV vehicle system model was implemented using the ADAMS command language. A key element of this model is a slip-sinkage model. This simulation model was validated using real-world data collected in a step validation fixture with sand and a variable-height curb. In a related effort, this project developed methods for using UGV sensor data to estimate variables and parameters needed for traction force prediction. The methods were evaluated using data collected from experiments with a PackBot traversing various deformable and non-deformable surfaces.

iii Summary Active vision systems have mechanisms that can actively control camera parameters such as position, orientation, focus, zoom, aperture and vergence (in a two camera system) in response to the requirements of the task and external stimuli. They may also have features such as spatially variant (foveal) sensors. More broadly, active vision encompasses attention, selective sensing in space, resolution and time, whether it is achieved by modifying physical camera parameters or the way data is processed after leaving the camera. In the active vision paradigm, the basic components of the visual system are visual behaviors tightly integrated with the actions they support; these behaviors may not require elaborate categorical representations of the 3-D world. Because the cost of generating and updating a complete, detailed model of our everyday environment is too high, this approach to vision is vital for achieving robust, real-time perception in the real world. In addition, active control of imaging parameters has been shown to simplify scene interpretation by eliminating the ambiguity present in single images. This document describes promising directions for research in active vision and possible applications of this research. It also discusses progress in experimental equipment for supporting this research and potential applications. Important research areas in active vision include attention, foveal sensing, gaze control, eye-hand coordination, and integration with robot architectures: Attention Selective processing of regions with restricted location, motion, or depth, is necessary for achieving real-time performance with limited resources. Although the design of the attention system is potentially highly complex, and will aaect the design of visual processing at all levels, little is known in detail about what this system should look like. Foveal Sensing A spatially-variant (foveal) sensor permits high resolution at the location of interest without the cost of uniformly high resolution. Issues that need exploring include the properties of unconventional \log-polar" coordinate systems speciically suited to foveal sensing, and the determination of visual features which v can be informative at low resolution to allow reliable selection of targets in the periphery. Gaze Control The alteration of imaging parameters to aid in the performance of visual tasks, or gaze control, is useful for many tasks, including image stabilization, overcoming a limited eld of view, gure-ground separation and range estimation. Gaze control is divided into two primary categories: Gaze stabilization and gaze change. The former consists of controlling the camera to maintain clear images of some world point that …

Research teams use common robots and machine learning to teach the robots outdoor navigation and locomotion skills.

This article describes the conduct of six evaluation experiments for the Perception for Off-Road Robotics program. Key distinctions of the testing methodology include conduct of the experiments by a group independent from the developers, unrehearsed experiments that provide little advance knowledge of the test courses, and blind experiments that do not allow the system operators to see the test courses until testing has completed.The article presents quantified, objective performance metrics for the systems evaluated. The basis for evaluation is 296 runs traveling 130 km in 110 hr. The results show significant progress over the course of the program, reducing the Emergency-Stops per kilometer by a factor of 22, reducing the uplink data volume per unit distance by a factor of 46 and the downlink data volume per unit distance by a factor of 3.At the end of Phase III, typical performance in desert terrain by the most reliable system achieved travel speed of 66 cm/s covering 90% of the distance in autonomous mode.

The DARPA Learning Applied to Ground Vehicles (LAGR) program is accelerating progress in autonomous, perception‐based, off‐road navigation in unmanned ground vehicles (UGVs) by incorporating learned behaviors. In addition, the program is using passive optical systems to accomplish long‐range scene analysis. By combining long‐range perception with learned behavior, LAGR expects to make a qualitative break with the myopic, brittle behavior that characterizes most UGV autonomous navigation in unstructured environments. The very nature of testing navigation in unstructured, off‐road environments makes accurate, objective measurement of progress a challenging task. While no absolute measure of performance has been defined by LAGR, the Government Team managing the program has created a relative measure: the Government Team tests navigation software by comparing its effectiveness to that of fixed, but state‐of‐the‐art, navigation software running on a standardized vehicle on a series of varied test courses. Starting in March 2005, eight performers have been submitting navigation code for Government testing on such a standardized Government vehicle. As this text is being written, several teams have already demonstrated leaps in performance. In this paper we report observations on the state of the art in autonomous, off‐road UGV navigation, we explain how LAGR intends to change current methods, we discuss the challenges we face in implementing technical aspects of the program, we describe early results, and we suggest where major opportunities for breakthroughs exist as LAGR progresses. © 2007 Wiley Periodicals, Inc.

A long-held dream for robotics researchers is the creation of vehicles that can move to a goal without human supervision, adapting as required to changing circumstances. While today’s ground robots are still far from achieving such complete autonomy, substantial progress has been attained. In this paper we describe the state-of-the-art in autonomous ground vehicle navigation as observed in the recently completed DARPA PerceptOR program, and we suggest new research directions where we see opportunities for leaps in performance.

rg ABSTRACT This paper describes a testbed and methods used in performing experiments and collecting quantitative data on the off-road mobility of two small ground robotic vehicles. The data is unique in the sense that it is: 1) unbiased, having been collected and interpreted by personnel independent of the vehicle developers, 2) locomotion-independent, since the same test procedures are followed regardless of whether the vehicle has wheels, legs, or tracks, 3) reasonably general, for the test range features a wide variety of terrain types including rock beds and mud pits, and 4) quantitative, in the sense that the results include measures other than pass and fail, such as voltage, current, and terrain ground truth. The paper reports on efforts to coordinate these testing capabilities with modeling and simulation for the purpose of predicting the mobility performance of a given vehicle on a given terrain. A series of basic to more complex dynamics models of a PackBot are used as a case study, along with their application to analysis of test results and formulating appropriate metrics for performance. Preliminary results in validating the model on steps, ditches, and slopes at the test range are presented.

The Department of Defense (DoD) is undergoing a transformation. What began as theoretical thinking, under the notion of a Revolution in Military Affairs (RMA) is now beginning to manifest itself in a "Transformation." The overall goal of the transformation described in Joint Vision 2020 is the creation of a force that is dominant across the full spectrum of military operations. The warfighting concept that will allow us to achieve Joint Vision 2020 operational capabilities is Network Centric Warfare (NCW). NCW is no less than the embodiment of an Information Age transformation of the DoD. It involves a new way of thinking about how we accomplish our missions, how we organize and interrelate, and how we acquire, field and use the systems that support us. It will involve ways of operating that have yet to be conceived, and it will employ technologies yet to be invented. NCW has the potential to increase warfighting capabilities by orders of magnitude, and it will do so by leveraging information superiority. A major condition to success is an infostructure that is robustly networked to support information collection, sharing and collaboration; which will require increased emphasis on sensor research, development and implementation. DARPA is taking steps today to research, develop and implement those sensor capabilities. The Multi-Body Control program is a step in that direction.

This report develops a systematic method, based on fractal gemmetry, for modeling natural terrain. The method consists of two main parts: reconstructing dense surfaces from sparse data while preserving roughness, and estimating the uncertainty of each reconstructed point. In earlier work, Szeliski developed stochastic ngularization techniques to reconstruct natural surfaces. We found that these methods did not provide sufficient control over the roughness of the reconstructed surfaces. We present a modified version in which a temperature parameter, determined as a function of the fractal dimension, plays a critical role in controlling roughness. Reconstructing dense, rough surfaces is seldom useful without assigning some measure of confidence to the surface points. This is particularly challenging for the reconstructed points. We revisit Szeliski’s approach of Monte Carlo estimation of uncertainty, and report quantitative accuracy results for both synthetic data and real range data.

This paper describes (1) a novel, effective algorithm for outdoor visual position estimation; (2) the implementation of this algorithm in the Viper system; and (3) the extensive tests that have demonstrated the superior accuracy and speed of the algorithm. The Viper system (Visual Position Estimator for Rovers) is geared towards robotic space missions, and the central purpose of the system is to increase the situational awareness of a rover operator by presenting accurate position estimates. The system has been extensively tested with terrestrial and lunar imagery, in terrains ranging from moderate—the rounded hills of Pittsburgh and the high deserts of Chile—to rugged—the dramatic relief of the Apollo 17 landing site—to extreme—the jagged peaks of the Rockies. Results have consistently demonstrated that the visual estimation algorithm estimates position with an accuracy and reliability that greatly surpass previous work.

The perception of sources of percussive sounds, as a function of variables that govern sound synthesis, was investigated. In the first set of experiments, perception of the source’s material was investigated. Two types of experiments were conducted. In the first, subjects judged similarity with respect to material and length. The sounds corresponded to modal vibrations of struck clamped bars. They varied in fundamental frequency and frequency-dependent rate of decay. Differences between sounds in both decay and frequency affected similarity judgments, with decay playing a substantially larger role. In the second type of experiment, subjects assigned sounds to one of four material categories. Decay parameters for each category were estimated and found to correlate with measurements reported in the literature. The second set of experiments investigated perception of contact location on the object. Subjects were presented with pairs of sounds produced by striking a string at two locations. Two strings with different fundamental frequencies were used. In one experiment, subjects judged the distance between the strike points, and in another, judged which point was closer to the point of attachment. The results confirm that subjects are sensitive to the interstrike distance and can judge which point is closer.

Contact sounds can provide important perceptual cues in virtual environments. We investigated the relation between material perception and variables that govern the synthesis of contact sounds. A shape-invariant, auditory-decay parameter was a powerful determinant of the perceived material of an object. Subjects judged the similarity of synthesized sounds with respect to material (Experiment 1 and 2) or length (Experiment 3). The sounds corresponded to modal frequencies of clamped bars struck at an intermediate point, and they varied in fundamental frequency and frequency-dependent rate of decay. The latter parameter has been proposed as reflecting a shape-invariant material property: damping. Differences between sounds in both decay and frequency affected similarity judgments (magnitude of similarity and judgment duration), with decay playing a substantially larger role. Experiment 2, which varied the initial sound amplitude, showed that decay rate—rather than total energy or sound duration—was the critical factor in determining similarity. Experiment 3 demonstrated that similarity judgments in the first two studies were specific to instructions to judge material. Experiment 4, in which subjects assigned the sounds to one of four material categories, showed an influence of frequency and decay, but confirmed the greater importance of decay. Decay parameters associated with each category were estimated and found to correlate with physical measures of damping. The results support the use of a simplified model of material in virtual auditory environments.