RRG Research Mobile Manipulator

Research

Mobile Manipulation

Motivation

Mobile manipulation offers a dual advantage of mobility offered by the platform and dexterity offered by the manipulator. The degrees of freedom of the platform also add to the redundancy of the system. These systems can be effectively used for a variety of tasks such as D & D, service robotics etc.

Objective and scope

In the Mobile Manipulator research Thread, UTRRG's objectives are

To develop the : kinematic, dynamic and stiffness models for the mobile manipulator system.
To develop performance criteria for mobile manipulation systems to integrate analytical models and sensory data for use in a decision making system.
To develop analytical and empirical task models for various end-effector tools for use as inputs to the mobile manipulation task-based control system.
To embed these results in OSCAR framework.

Approach

Our approach is different from traditional artificial intelligence techniques that aim at full autonomy, as it acknowledges the use of pre-determined models along with sensory data and human input for conflict resolution. We feel that this approach combined with learning techniques for un-modeled parameters can offer a comprehensive solution towards full autonomy. Its key components are:

Manipulation and Mobility Software: OSCAR is an object-oriented framework that provides the building blocks for developing advanced manipulator control applications that support teleoperation, automation, and human interaction. Central to OSCAR are decision making algorithms that use a set of 50 different performance criteria to prioritize the operation of the manipulator. These criteria range from simple constraint, to kinematic, dynamic, compliance, obstacle avoidance, and motion planning. The combination of the criteria used and their relative importance allows customization of the manipulator behavior to address a specific task requirement. Examples of task requirements are: ability to exert force in a specified direction, minimize vibration, maximize accuracy, etc. The basic analytical research used by OSCAR has been done over the past 25 years and is being aggressively continued, especially in the area of task-based performance specification, obstacle avoidance and motion planning, force control, and mobility. At the basic level, OSCAR also has robot independent software components for kinematics, dynamics, and stiffness computations.

Our partners at INL have developed an object-oriented mobile system control software framework that adapts easily to various robot geometries and to sensor suites. The entire framework (complete with all behaviors and associated autonomous control) is easily ported by simply editing a few parameters (i.e., robot length, width, maximum speed) in a script file. Moreover, the system allows the robot to recognize what sensors it has available at any given time and adjust its behavior accordingly. This research will 2 integrate OSCAR with the INL framework to provide a comprehensive software package for mobile manipulation that is independent of the robot geometries, kinematic configurations, sensor suites, and a variety of end-effector tools.

Back to Top

Testbed

System: A modular manipulator testbed is developed for demonstrations and experimentations. It consists of the Powercube^TM actuator modules manufactured by Amtec GmbH, Germany mounted on top of IROBOT's ATRV2 mobile robot. The mobility intelligent kernel and the Manipulator intelligence (OSCAR) are on the onboard LINUX PC. The manipulator is commanded through CAN-USB mini.

Figure 1. The testbed (PowerCube Manipulator mounted on ATRV platform)

Sensing:

Force Torque Sensor: To sense contact forces, a wrist-mounted multi-axis ATI FT 30/100 Force/Torque (F/T) sensor is used. This measures all six components of the end-effector forces and torques. The axial, transverse and torsional load capacities are 300 lbs, 150 lbs and 600 lb-in respectively. The maximum attainable data transfer rate is 600 Hz with the controlled system. Accuracy (including manipulator vibration) during operation is 0.25 lbs.

Positioning and Mapping Sensors: The testbed is equipped with laser scanner and sonar sensors for building the world map and to aid positioning. Basin odometric and inertial sensors are used for primary positioning of the robotic system in the workspace.

For correcting the position of the platform ARCSecond's indoor GPS (iGPS) is used externally.

Software: The operational software used for developing the mobile manipulator applications is a heterogeneous software consisting two entirely different software applications OSCAR (developed by UT) and the mobility intelligence kernel (developed by INL).

Research Results

Simulations

	This is a simple demo of the first phase of the project The demonstration of an object retrieval. The object is identified using iGPS. (~10 MB)
	Real-time motion planning of PowerCube manipulator The demonstration of iGPS-PowerCube Integration. The iGPS sensor is actively tracked by the PowerCube manipulator. (~1.2 MB)
	A door opening demo A door opening demonstration. the iGPS system is used for the locating the door knob and also for real-time positioning of the mobile manipulator (~2 MB)

Publications

Cox. D. J, Pryor. M, Cetin. M, and Tesar. D, June 1999, "Experiments in Cooperative Manipulation for Dual Arm Robotic Operations", Proceedings of the 10th World Congress on Theory of Machines and Mechanisms [Abstract] [Full-text PDF]

Hester. R. D, Cetin. M, Kapoor. C, and Tesar. D, May 1999, "A Criteria-Based Approach to Grasp Synthesis", Proceedings of the IEEE Conference on Robotics and Automation , Vol. 2, pp. 1255-1260 [Abstract] [Full-text PDF]