Research & Development

Snake Arm

This effort serves to further the efforts of the National Center for Defense Robotics to investigate and develop robotic capabilities for use in current and future robotic combat casualty care and future needs in Autonomous Combat Casualty Care. This project is intended to further the knowledge base for robotic capabilities on the battlefield, in supporting first responders, and in reducing the risk to military personnel. The integration of a zipper mast and a snake arm onto an ARES UGV platform seeks to address two needs related to autonomous combat casualty care:

  • Enable remote fine-scale interactions with patients by integration of a highly dexterous manipulator on a UGV
  • Quickly locate potential patients despite low lying battlefield obscurants such as brush, grass, and walls by deploying sensors at sufficient altitudes to look into windows and over obstacles

Combining these two capabilities onto a single UGV furthers the knowledge base for robotic capabilities.

Carnegie Mellon University has built Snoopy II, part of a family of hyper-redundant robot arms (“snake arms”) in the Biorobotics Lab. Essentially, they are a sequence of actuated universal joints stacked on top of each other where each joint provides two degrees of freedom. Snoopy 2 has 6 such joints, for a total of 12 degrees of freedom, and is 56cm (22in) in length.

The TeraBot and Snoopy II are interfaced through a single computer (TS-7800) which accepts JAUS messages and performs the kinematic calculations to provide end-effector operation. The power and signal lines for the camera mounted on Snoopy II are passed through the TeraBot’s internal wiring, while the power and communication lines for Snoopy II itself have been mounted externally to the TeraBot. Communication with the TeraBot is done over an RS-232 serial port while the snake arm is interfaced using a USB to RS-485 converter. Each manipulator uses its own custom protocol to send and received messages with the TS-7800, which communicates information with the OCU using JAUS messages. The TS-7800 is connected to the ARES platform through one of its payload ports, which supplies both communications and power.

We developed combined kinematic models of the TeraBot and Snoopy II manipulators. The following table lists the Denavit-Hartenberg (DH) parameters for the combination of the Oceaneering TeraBot with the CMU Snoopy II given the dimensions of the mount to attach Snoopy II to the TeraBot, the height of the ARES platform, and the assumption that Snoopy II is mounted distally along the TeraBot’s second link. The intermediate coordinate systems used were chosen such that all positive rotations about horizontal rotation axes cause the arm to lift up, and all positive rotations about vertical rotation axes cause the arm to rotate to the left.

  • Link diameter: 47 mm (1.85 in)
  • Module length: 96 mm (3.78 in)
  • Joint travel: (+/-) 55 degrees max.
  • Bend radius, min. (on joint CL): 104 mm (4.09 in)
  • Gear ratio, overall: 2485 to 1
  • Angular resolution: 80,000 counts per rev. at output (0.0045 degree)
  • Joint backlash: 0.06 degree
  • Joint speed, theoretical no-load: 0.11 rev/sec or 2.8 sec full travel (+/-55 degrees)
  • Module mass: 230g (0.507 #)
  • Joint torque, continuous: 4.5N-m (39.9 #-in) @0.24 A (90C)
  • Joint torque, peak: 10.6 N-m (94.0 #-in) @0.56 A, 12V
  • Cantilever capability ( # modules @ cont. torque rating): 6.5
  • Bus wiring: Power + RS485 bus: (2) #20 + (2) #30

Inverse velocity kinematics were developed and implemented for the combined kinematics model. A 3D graphical simulation of the combined TeraBot and Snoopy II manipulators was developed to aid in the testing and implementation of the combined inverse kinematics. Three modes for operating the end-effector were implemented in simulation. The operator can simultaneously specify: the change in end-effector position and orientation, only the change in position and allow the orientation to vary, or can specify the orientation while holding the current position.

These three end-effector modes are implemented by computing the Manipulator Jacobian of the combined arms during each iteration of the control loop. The Jacobian is computed from the individual transformation matrices of each joint for the current configuration. Singular value decomposition is used to determine a pseudo-inverse to the Jacobian, allowing for the computation of required changes in joint angles to attain a desired change in end-effector position and/or orientation. This method provides a least-squares approximation for the solution to determining needed changes in joint angles and as such can produce undesired movements of the end-effector when near configurations of low manipulability.

Finally, a new widget was built for SURC to handle manipulator information. The widget allows the operator to select the desired control input method. Both end-effector and joint by joint control are supported. The TeraBot manipulator is treated as a separate manipulator than the Snake Arm in this implementation. Though a single manipulator control scheme was designed, it turned out to be too computational intensive for our embedded single board computer to effectively handle. The operator can quickly switch between arms by a mouse click. The joystick mappings automatically change to support the selected joint control mode.

© 2008 Copyright Applied Perception, Inc. - A QinetiQ North America Company | 220 Executive Drive, Suite 400 Cranberry Township, PA 16066