The push at RRG has always been towards modularity in robotic systems.  Modularity is essential if a system is to be reusable, low cost and of increasingly higher performance by continuous updates [Tesar, 1997].  Modular components can easily be reconfigured to meet changing operational requirements and they enable quick repair and maintenance of the robot.  Two modular robotic manipulators developed at RRG are the ALPHA (Advanced Lightweight Prototype High-performance Arm) and MARS (Modular Arc-welding Robot for Shipbuilding) manipulators.

Conceptually, the ALPHA manipulator is a high payload/weight, modular, and extremely accurate all-revolute, 7-DOF serial robotic manipulator designed under contract for the Defense Advanced Research Projects Agency (later renamed the Advanced Research Projects Agency or ARPA) [Marrs and Tesar, 1997].  The requirements for the ALPHA were developed from analyzing different types of operations that went beyond low-value tasks such as welding.  It was designed to perform a broad spectrum of industrial (such as precision operations like die production, air frame assembly and fully integrated manufacturing cells) and military applications.  To achieve high precision under load, the ALPHA manipulator was designed to be structurally very stiff and sources of dynamic shock such as backlash and static friction were reduced.  At the end of the design and development phase, the RRG demonstrated the lightweight, sensor-based manipulator built up of advanced shoulder, elbow and wrist modules and commanded by a sophisticated control package.  A conceptual drawing and a list of the target specifications are listed below.

The MARS manipulator is a 6-DOF, low-cost, lightweight system of actuators, links and networked actuator controllers that was designed for arc-welding in shipyards designed under contract for ARPA [Grupinski and Tesar, 1997].  Due to the conditions in shipyards, the components comprising the MARS manipulator were designed to be impervious to dirt and moisture and to withstand abuse.  The MARS manipulator was mounted on a gantry and was used to weld stiffeners to large panels.  The target specifications for this robot are given below.

Most recently, the RRG was involved in the design of a rugged manipulator to be used for decomissioning and dismantlement (D&D) tasks where the foremost use of robotics technology for such tasks is the elimination of human on-site involvement.  The robot will be used to transport, sort and pacakage heavy metals in an unstructured, obstacle-ridden environment for long periods of time, implying that the robot must be rugged with respect to accidental shock and misuse and it must be able to exert and withstand large forces, which further implies large deformations.  The robot must therefore be equipped with stiff actuators and links.  Specifically, the robot target specifications were: 110 kg (~250 lb) payload capacity, a reach of 2.03 m (80 in) and an end-effector speed of 0.91 m/sec (36 in/sec).  To date, the design of the elbow module for this rugged, modular robot has been completed and is ready for manufacture.  This module has peak and nominal output torques of 850 N-m (7500 in-lb) and 170 N-m (1500 in-lb), respectively, and an output speed of 30 rpm.

A serial robotic manipulator can basically be thought of as an electromechanical system consisting of two basic building blocks: links and actuators.  Currently, many tools exist to model serial manipulators at the system level as well as actuators independently from one another with a high level of detail and accuracy.  Although it is well known that the properties of the actuator define the character of the manipulator that houses it, a limited amount of work has been done to treat the central issue of modeling the combination of the two.

Past research in the area of robotic manipulator design has been to develop system level design metrics based on geometry and dynamics.  Metrics for estimating the acceleration capability, force capability, kinetic energy, compliance/stiffness, accuracy, etc. have been developed at RRG.

"Design of Mechanical Properties for Serial Manipulators." (PDF)
"Task-Based Decision Making and Control of Robotic Manipulators." (PDF)
"An Emperical Approach to Performance Criteria and Redundancy Resolution." (PDF)

However, to achieve higher performance and better design of robotic systems, the effect the actuator's parameters have on the capabilities of these highly complex systems have to be understood.  Significant work has been done at RRG to design EM actuators that can be used for an array of applications including robotics.  Based on this knowledge and experience, many design tools have been developed at the actuator level that can be used for system level design as well.

Actuator Design and Architecture
Decision Making and Performance Criteria

The current research goal in the area of serial robotic manipulator design at RRG is to improve the science of the influence of the actuator parameters on the system capability of these systems leading to higher performance and system level configuration management.

The process begins by choosing a kinematic configuration that has some desired characteristics such as reach and dexterity.  Next, parametric modeling of the EM actuator is performed allowing the designer to calculate what the status of a particular actuator design is (i.e. weight, inertia, torque, etc.) based on some parameter(s) the designer is free to choose (i.e. material properties, dimensions, etc.).  Parameters such as dimensions and material properties are known as design parameters because they can be varied by the designer while parameters such as weight and torque are known as performance because they dictate the performance of the actuator.  Intermediate parameters also exist and they are groupings of several design parameters.  One example of an intermediate parameter is the gear reduction ratio because it depends on several dimensions of the gear train.  The performance parameters are typically non-linear functions of the design and intermediate parameters.

Now that the kinematic and dynamic properties of the robotic manipulator have been established, global performance/design maps can be constructed via some performance/desing measures or metrics.  These metrics are used to calculate what the properties of a particular manipulator design are.  Some examples of these metrics are the Inertia Frobenius Norm (IFN) which measures the amount of kinetic energy in the system, the achievable end-effector acceleration capability which measures the robot's ability to accelerate its end-effector from rest and the achievable end-effector static load which measures the robot's ability to apply or resist an external static load.  Using these metrics, a particular design can be evaluated in a global sense and evaluating many designs yields the global performance/design maps for IFN, acceleration capability, force capability, etc.  The different designs are achieved by changing the properties of the actuators.

The final steps in the design process involve developing constraint equations based on physical limitations on the actuators such as motor speed, gear teeth strength, etc. and on the system.  Optimization techniques are employed to select the actuator parametes that yield the best design given the constrains.  This last step is known as system level configuration management.

Ambrose, R. and Tesar, D.  "Design, Construction and Demonstration of Modular, Reconfigurable Robots."  Disseration, UT Austin, 1991.

Butler, M. and Tesar, D.  "An Applications-Based Assessment of Present and Future Robot Development."  Thesis, UT Austin, 1991.

Hill, B. M. and Tesar, D.  "Design of a Serial Three Degree-of-Freedom Shoulder for Modular Robots."  Thesis, UT Austin, 1993.

Hernandez, E. and Tesar, D.  "Compliance Modeling for General Manipulator Structures."  Disseration, UT Austin, 1996.

Browning, G. and Tesar, D.  "The Physical Significance of Performance Criteria for Redundant Manipulators."  Thesis, UT Austin, 1996.

Hill, B. M. and Tesar, D.  "Design of Mechanical Properties for Serial Manipulators."  Dissertation, UT Austin, 1997.

Marrs, M. R. and Tesar, D.  "Design of an Advanced, High-Precision, Seven Degree-of-Freedom Modular Robotic Manipulator."  Thesis, UT Austin, 1997.

Tesar, D.  "Where is the Field of Robotics Going?"  White Paper, UT Austin, 1997.

LeGoullon, A. P. and Tesar, D.  "Configuration Management of Robotic Workcells."  Thesis, UT Austin, 1997.

Legault, J. M. and Tesar, D.  "A Complexity Management Framework for Open Agile Manufacturing Systems."  Thesis, UT Austin, 2000.

Turner, C. J. and Tesar, D.  "A Criteria Development for Actuator Resource Management."  Thesis, UT Austin, 2000.

Pehl, J. E. and Tesar, D.  "Analysis for Design of Serial Manipulators."  Thesis, UT Austin, 2002.

Ashok, P. and Tesar, D.  "Design Synthesis Framework for Switched Reluctance Motors."  Thesis, UT Austin, 2002.

Gloria, C. and Tesar, D.  "Parametric Modeling and Design Synthesis for Electromechanical Actuators."  Thesis, UT Austin, 2004.

OSCAR v2.0: Online reference manual for Operational Software Components for Advanced Robotics (OSCAR) C++ libraries. OSCAR contains libraries for sensing and control

RoboWorks: Graphical package for simulating robot control

RRG Simulations Website: A page maintained by UTRRG on simulations for engineering education. Contains good examples of application development for manipulator control