Cregg K. Cowan

Determining the camera and light source location for a visual task

Techniques for calculating the three-dimensional region of acceptable locations for vision sensors and light sources are discussed. An integrated method to position both a camera and a light source based on the requirements of a given vision task is described. The method also selects the appropriate lens aperture. The proposed techniques require only a small amount of computation (approximately 1 minute of elapsed time on a time-shared VAX 8600). The proposed approach is illustrated with an example object and experimental results are described for that object.<<ETX>>

Model-based synthesis of sensor location

A system that automatically generates the possible camera locations for observing an object is described. The approach, which uses models of the object and the camera, is based on meeting the requirements that the spatial resolution be above a minimum value, all surface points be in focus, all surfaces lie within the sensor field-of-view, and no surface points be occluded. Each sensing requirement is converted into a geometric constraint on the sensor location, from which the three-dimensional region of viewpoints that satisfies that constraint is computed. The intersection of these regions is the space where a sensor may be located.<<ETX>>

Edge-based placement of camera and light source for object recognition and location

The selection and placement of cameras and light sources for a vision task is an essential step in both autonomous visual sensing and standard industrial applications. A technique is presented that determines the three-dimensional region of light-source locations such that one or more specified object edges will be detected with a given edge operator. The method uses a task description that includes a list of object edges and the edge operator to be used in order to derive constraints on image contrast, surface irradiance, and light-source location. The combination of this method with prior results on camera and light-source placement can be applied to object recognition and location tasks.<<ETX>>

Understanding object configurations using range images

The authors describe techniques that generate multiple interpretations from dense range images of piles of unknown objects and methods that use physical law, such as object stability, to rank the interpretations. Each of the interpretations completely accounts for the observed range data, but the interpretations differ in the ways visible portions of objects are extended into the occluded portions of the scene. Experiments with 100 range images indicate that the techniques are fairly robust when the scenes consist of shapes that are approximately prismatic or cylindrical. These techniques are based on novel approaches in several key areas, including explicit use of sensor geometry, generic shape models to synthesize scene descriptions, spatial-reasoning techniques that incorporate knowledge about the laws of physics, direct estimation of the physical properties of the objects in the scene, and detection and refinement of descriptions of approximately planar or cylindrical surfaces. >

Automatic light-source placement for detecting object features

The selection and placement of cameras and light sources for a specific task (e.g., locating a part in a tray or inspecting an object) is one of the most important steps in creating a successful vision system, because obtaining high-quality images can greatly simplify the vision algorithms and improve their reliability. We will describe techniques that use a visual task description stated in terms of features to be detected, and derive a range of light-source locations that satisfy the task requirements. In particular, given a task description that specifies particular object edges to be detected with a given edge detector (e.g., a Sobel edge operator), our techniques determine the constraints on light-source location such that the edge is detected.

Application of micro-robots for building carbon fiber trusses

Using diamagnetic micro-manipulation, micro-robots (~10 mm in size) with specialized end-effectors are able to grasp and manipulate small light weight (~mg) carbon fiber elements to construct macroscopic cubic carbon fiber trusses (~1000 parts and 300 mm long) that can achieve mechanical crush performances surpassing those of commercially available aluminum honey combs. Employing low cost micro-robots and end-effectors allows the implementation of both parallelization and new functionality without the need for expensive fixed tooling. We demonstrate this concept by modifying the micro-robot end effectors to build carbon fiber octet trusses and carbon fiber trusses with integrated electronics.

Optimal control of diamagnetically levitated milli robots using automated search patterns

This paper reports early results of optimal control using diamagnetically levitated robot (DLR) systems. Levitated robots, typically 1-4 mm in size, are driven locally by printed circuit boards (PCBs) with zero friction and near-zero hysteresis. As previously reported [1, 2], these properties give the levitated system excellent repeatability at the micron and submicron levels for both positioning and trajectories. Highly repeatable systems are generally good candidates for optimal control because variability in identifying and exploiting optimized control parameters are minimized. In the data reported here, levitated milli robots are characterized, and we present experiments showing a 5× reduction in oscillations using a debouncing routine that is manually optimized in time. Manual optimization is effective in some cases, but due to the multi-degree-of-freedom (DOF) aspects of levitated robots, a more robust self-tuning method is desirable. Toward this goal, we describe experiments using automatic optimization of robot motion using a computerized search-and-test routine that varies PCB trace currents and selects optimal currents based on automatic measurement of motion error. The automatic optimization using this method has shown settling times to micron and submicron levels that are 20-40 times faster than simple bang-bang control of equilibrium currents. In one test reported here, 150-μm moves were demonstrated at 90-nm rms error with 15-ms move times.

Automated 2D micro-assembly using diamagnetically levitated milli-robots

In this article, we demonstrate the application of diamagnetically levitated milli-robots for the 2D micro-assembly of 10-μm polymer microspheres and other silicon microfabricated parts. By using an optical microscope for feedback (imaged at 27 Hz), we are able to demonstrate long-term open-loop stability (up to 78 hr) and sub-micron stability of the levitated micro-robots. Furthermore, due to the low hysteresis and high compliance in the magnetic drive of the milli-robots, we are able to directly use the milli-robots in conjunction with machine vision as a force sensor. Soft polymer-based end effectors are used for the micromanipulation of parts and show modest reliability of pick (>70%) and high reliability of place (>99%) that is insensitive to the pick surface material. Finally, we implement autonomous micro-assembly from randomly deposited microspheres into ordered arrays.

Multi-agent systems using diamagnetic micro manipulation — From floating swarms to mobile sensors

Multi-agent robotic systems on small scales typically use, or envision using, small mobile robots. This paper takes a broad-brush look at a relatively recent development in milli and micro robots, Diamagnetic Micro Manipulation (DM3) systems, in a multi-agent context. We report various multi-agent operations, such as multi-agent part manipulation, robots-acting-on-other-robots for enhanced functionality, and robot-supplied local feed, to assess the state of the art of this new type of system. In addition, we explore and analyze the multi-agent implications of known DM3 properties and earlier single-agent results to motivate future research in this area. As reported here, some of these multi-agent implications include mobile high-precision sensor transducers, near-field communications channels using robots as both transmitters and receivers, and novel swarm architectures.

Visual registration for robotic operations on space shuttle tiles

Refurbishing the thermal-protection tiles on a space shuttle before each mission is a lengthy and labor-intensive process. A mobile robot is being developed (described elsewhere) to perform two of the required maintenance operations on the bottom side of the shuttle: (1) injection of a hydrophobic fluid, to prevent tiles from absorbing water, and (2) visual inspection, to detect anomalous tile conditions. Both operations depend on precise positioning of the robot end effector with respect to each tile. We describe our method for precise visual registration. The technique first detects the edges of the tile (whose approximate shape and dimensions are given from CAD data) and then uses correspondence between visual features in the post- and pre-flight images to improve the registration accuracy. Results on actual tile images are presented.