 |
|
Brief: The RoboSight project started in early 2008 as a result of meeting Guy Paillet of General-Vision at the AAAI conference held in Vancouver, B.C. in late 2007. General-Vision is the inventor of the CogniMem silicon neural network chip. RoadNarrows Robotics subsequently established a partnership with Recognetics, the distribututor of the CogniMem and General-Vision. Carnegie Mellon University designed the very popular CMUcam small smart camera - a µprocessor based module. RoadNarrows decided that the robotics market could bear another smart camera for robotics and embedded applications but with an alternative approach to vision processing with an artificial neural network rendered in silicon. |
 |
|
Brief: The SkewlZone™ project scope aims to develop a generalized processing and sensor pack complete with open-source software and curricula for using legged robots as a platform in Science, Technology, Engineering, and Mathematics (STEM) courses. The end goal is to commercialize the middleware SkewlZone™ Brain and Sensor Pack. The Brain Pack consists of processing boards, plug-in sensors, wireless communication, and mechanical hardware for attachment to commercially available legged robots. It serves as a value-adding layer between the low-level operations of a robot platform and the high-level software. Educators, researchers, and customers will be able to migrate their Brain Pack from one robot to another, allowing the re-use of software, curriculum, and hardware. The Brain Pack provides sufficient sensory feedback to close the loop on the mechanical control of the legged platform. Provided sensors, such as foot and hand haptics and an inertial measurement unit, are critically important for legged robots. In a similar philosophy, open-source robotics software, such as Pyro and Tekkotsu, abstract the details of the robot’s hardware. Higher-level cognitive algorithms, such as path-planning, vision, and behavior based systems, can be readily developed or reused. |
 |
|
Brief: The main focus of Fusion (besides being fun) is to add value to the robots sold by RoadNarrows Robotics. Fusion comes as an open source software package written in python. Its architecture is built around a core controller (Reactor) and a set of utility subpackages with pluggable virtual vRobots, vBrains, vSensors, and visualizers. The physical robots and hardware are controlled over wire(less) messaging and data streams. To date, the Khepera II, Hemisson, and KHR-2 robots are supported with various vBrains. The vBrains are the control methodologies and behaviours associated with a particular robot. Different vBrains can be plugged into the same vRobot. The next version of Fusion will be a complete rewrite from the lessons learned. The underlining fabric will be Gluon, a messaging interface with a defined set of operations such as (un)registering, attribute advertising, push/pull data streams, etc. With Gluon, the plug-in modules can be written to a standard set of interfaces. Gluon will provide the underlining “magic”, and with Gluon, a vRobot may, for example, control a physical robot, the host where the Fusion Reactor is executing while a visualizer plugin remotely plugs-in to the Reactor from a different node connected by a network. |
 |
|
Brief: The 1049 project evaluated state of the art small hydrogen fuel cell technologies for mobile robotic platforms. Hydrogen fuels cells hold promise of tremendous portable power desperately needed in mobile devices. Moreover, introducing fuel cells to engineering students exposes them to new, clean energy possibilities. RoadNarrows evaluated the products from Horizon Fuel Cell Technologies. The small fuel cells were well engineered and had good potentential, however, hydrogen storage was felt to be still somewhat problematic. High pressure hydrogen can present hazards when either used directly as the source feed to the fuel cells or used for replenishing metal hydride reabsorption. In fact, because of the difficulties associated with hydrogen storage, the U.S. Department of Energy initiated the “National Hydrogen Storage Project” in 2003 involving many universities, companies, federal laboratories and international partners to actively engage in hydrogen storage research. Unofrtunately, the fuel cells were also quite expensive and not competitive with existing battery technologies. Therefore, given the current state, the 1049 project has been deferred until the technologies mature. |
 |
|
Brief: The OrcBoard project had two major impacts on RoadNarrows. First, we were collaborating with M.I.T., already a very good customer. The OrcBoard was the brain-child of Ed Olsen, a then engineering graduate student at M.I.T. The OrcBoard is an open reference, hardware, firmware, and software general-purpose controller board for robotics. It has very nice peripherals plus a remote control interface with a Java run-time environment. The M.I.T Maslab used the OrcBoard to introduce engineering students to robotics. Second, until 2005, RoadNarrows had manufactured only simple electronics. We needed to gain expertise, not only in design, but in production compentencies. RoadNarrows performed some minor design changes and produced several dozen OrcBoards for Maslab classes starting in the fall of 2006 and 2007. However, the market for general-purpose robotic embedeed systems and controllers is very crowded and competitve. Therfore, it was decided that there would little business advantage for RoadNarrows to continue producing the OrcBoard. |
 |
|
Brief: The Pedxing concept was born from the hard business lessons learned at RoadNarrows from selling high-quality R&D robotic platforms and the invaluable feedback from customers integrated with emerging technologies of networking, power and battery management, processing performance, and inter-operabilities. This project served as Roadnarrows' entry into robotics manufacturing. In 2005 RoadNarrows applied for an National Science Foundation grant to fund the Phase I efforts to prototype and evaluate a Power over Ethernet (PoE) centric modular robotic platform. Although there were many positive responses from the reviewers, the application was not accepted. However, RoadNarrows strongly believes there are still very good market potentials for Pedxing. The combination of moderate costs, performance, extensibility, and standard operation packaged in a clean industrial design will always have market appeal. |
 |
|
Brief: The iixing (pronounced “eye-zing”) concept arose from concepts of dataflow language models, the emerging mult-processors and fast peer-to-peer networks, and the new grid computing architecutres that harness inter-connected heterogeneous processors and OS's. The mammal visual cortex is a massively parallel dedicated "wetware" construct that seemed an ideal match to dataflow architectures, hence iixing. The business model was to produce a suite of software and, perhaps later, dedicated hardware for commercial vision processing applications. After a set of software was written to validate the iixing approach, a project checkpoint determined the resource required would exceed RoadNarrows expected growth, so the project was deferred. |
 |
|
Brief: The XCogSci project started before RoadNarrows LLC was formed. It was an out growth of a general-purpose Genetic Algorithm (GA) progam call Gaia written by Robin Knight. As the area of GA evolved (no pun intended) into the more expansive concepts of Evolutionary Computation and as the XML standards solidfied, it was realized that a more general framework was needed, one that was open and extensible, similar to the MathML and SVG specifications. Gaia, transformed into XClotho, would validate the efficacy of XCogSci.EC (a subset of XCogSci). However, as extremely interesting as this area of research is and the possible value added to the research community, little revenue from such an effort could be expected. |