Metrology Referenced ROving Accurate Manipulator (MR ROAM), 10-R8019

Printer Friendly Version

Principal Investigators
Glynn Bartlett
Paul Hvass

Inclusive Dates:  01/01/09 – 01/01/10

Background - Modern industrial robots are generally monolithic assemblies, either floor-mounted or gantry-mounted, whose accuracy is largely determined by the error stack-up through each joint and a link from a ground reference to the robot's end effector. The fact that the systems rely upon a physical link to the ground reference means that there is limited robot market penetration for large-scale applications or for operation at remote sites. However, the recent availability of large-scale metrology systems that provide real-time, non-contact position data (e.g., Nikon Metrology's indoor Global Positioning System [iGPS] or motion capture technology) enables a signal-based link to ground reference that can replace the physical link. The robotics industry is poised for rapid expansion using off-the-shelf manipulators on mobile platforms, in conjunction with an external metrology system, for large-scale and remote site tasks where accuracy and repeatability are required over large working envelopes.

Approach - SwRI has demonstrated a Metrology-Referenced ROving Accurate Manipulation (MR ROAM) capability to serve both military and industrial clients. For this project, SwRI conducted a feasibility study to accomplish the following objectives:

  • Assemble and integrate a small-scale MR ROAM system from pre-existing and off-the-shelf components, including a Pioneer AT3 skid steer platform, Robai Cyton Alpha 7 degree-of-freedom manipulator, and Vicon motion capture system, all on a wireless network with Windows-based PCs for control (refer to Figures 1 and 2).

  • Implement a feedback control loop from an independent metrology device (i.e., Vicon) to provide base pose updates to the controlling software.

  • Quantify system and component-level contributions to accuracy and repeatability executing a 3D-path plan using the motion capture system for feedback and data acquisition.

  • Analyze the data to test this hypothesis statement: A metrology referenced mobile manipulation system can achieve 6 mm repeatability performing a dynamic material deposition task over a workspace that is an order of magnitude larger than its manipulator workspace.

To assemble MR ROAM, the manipulator arm was mounted forward of the platform steering axis on the centerline of the platform. A lightweight aluminum frame was constructed to hold a set of reflective fiducials so that the motion capture system could track the base pose of the manipulator. A second set of fiducials was mounted to the end effector to track its position for data acquisition. The motion capture system provided base pose updates to the controller at 120 Hz over a TCP/IP socket. A C++ program was written to control the system by leveraging the Gold version of the Agile Planet Kinematix™ software development kit. Because of difficulties in keeping the 9-DOF inverse kinematic solution within the joint ranges for the robot, the control was partitioned into two separate control loops, with commands going to the platform and manipulator individually. The control software ran on a Windows™-configured laptop, issuing commands to the robot over a wireless connection at a 40 Hz update rate. For the accuracy and repeatability testing the manipulator had no payload.

Figure 1. MR ROAM Test Setup

Figure 2. Screen Capture of an Example Path Plan

  Figure 1. MR ROAM Test Setup



Figure 2. Screen Capture of an Example Path Plan


Accomplishments - Empirical tests were conducted to quantify the repeatability and accuracy of the system. End-effector accuracy and repeatability were computed at 20 mm waypoint intervals along a curved 3D path plan. System repeatability, taken as the average ± one standard deviation, was better than the 6 mm threshold established for feasibility in low accuracy tasks like painting and paint stripping (at a speed of 40 mm/sec). However, system accuracy was inadequate due to compliance in the Cyton Alpha manipulator. In an effort to improve system accuracy, increase system scale, and implement state estimation algorithms, a second phase of MR ROAM has been undertaken during calendar year 2011.

MR ROAM Phase 2 - MR ROAM 2 is a 10 DOF mobile manipulation robot composed of a Vetex omni-directional platform, 7 DOF Motoman manipulator, and a Nikon Metrology iGPS system for position/orientation feedback (Figure 3). To date, MR ROAM 2 is performing better than expected. A final report will be available by the end of the calendar year detailing quantitative outcomes.

Figure 3. MR ROAM 2 following a curved path plan on an aircraft wing.


          Figure 3. MR ROAM 2 following a curved path plan on an aircraft wing.


 

2010 Program Home