Force/Torque Sensing Applied to Industrial Robotic Deburring

J. Norberto Pires, John Ramming, Stephen Rauch and Ricardo Araujo

Abstract

Force/Torque sensing is very important for several automatic and industrial robotic applications. Basically, if precise control of the forces that arise from contact between tools and parts is required to successfully complete the automatic task, then a force/torque sensor is needed along with some force/torque control technique. In this paper we focus on force/torque sensing aspects applied to industrial robotic tasks. Concentrating on a particular type of force/torque sensors, we demonstrate how to use them and how to integrate them into force/torque control applications using robots. Finally, an industrial application is presented where force control was fundamental for the success of the task.

Introduction

Due to their enormous flexibility robot manipulators are good examples of equipments for Flexible Manufacturing Systems. In fact flexibility is the major reason for robot utilization and popularity in the actual manufacturing plants. In this framework, the majority of the robot tasks require contact with the surrounding environment, i.e., in the process of fulfilling the task the robot tool interacts physically with the working objects and surfaces. That interaction generates contact forces that should be controlled in a way to finish the task correctly, not damaging the robot tools and working objects. Those contact forces depend on the stiffness of the tool and working objects/surfaces and should be properly controlled. The option for a particular control technique depends on identifying if:

1. The contact forces should be controlled to achieve task success, but is sufficient to keep them inside some safety domain: Passive Force Control.
2. The contact forces should be controlled because they contribute directly to the success of the task: Active Force Control.

In the first case, contact forces are an undesirable effect of the task and it is generally sufficient to keep them inside some safety domain. They are not necessary for the task, so usually the strategy used is adding flexibility to the endeffector with the object of damping all the possible impacts and increase the tolerance to positioning errors, complemented with detailed and careful planning of flying trajectories and object approach. There are many solutions in the market to add flexibility to the end-effector, and in fact this is currently the standard approach in industry.

In the second case, the contact forces are necessary to finish the task correctly, i.e., controlling the contact forces making them to assume some particular value or more generally to follow some force profile is part of the task.

In this paper, we focus on force/torque sensing and in the technical problem of adding force/torque sensing and active force control capabilities to an industrial robot manipulator. With that task in mind we will use an industrial robot with a wrist force/torque/acceleration sensor attached, and a personal workstation for programming, monitoring and control. Actual industrial robots controllers are closed controllers, not allowing access to motor and sensor signals even to the advanced user. Nevertheless, some of them have good programming environments with powerful tools and libraries. Because of that, in a way to explore them, we use a distributed and object oriented software architecture, developed by us, enabling users to add new functionalities to the original control system using also their available capabilities.

Force/Torque Intelligent Sensing

There are several ways to measure force but practical implementations of force/torque sensors rely basically in two methods:

1. Equilibrium condition between two forces, being one known and working as reference to obtain the other.
2. Determination of motion parameters imposed on a known mass by the unknown force.

Method 2 above is not simple to use, and in fact it is very hard to build sensors based on it, being only used, to the best of my knowledge, on astronomy and nuclear/atomic applications. First method is then the most used, although direct conversion between force and an electrical signal is not possible. There are several physical principles that can be used to obtain an electrical signal from the unknown force: strain gauge, magneto-elastic, piezoelectric, piezo-resistive, electromagnetic, etc.

For industrial robotics applications force/torque sensors are usually placed near the working tool, generally in the manipulator wrist. This means that the sensor must be reasonably small, built in several dimensions to adapt to different robot bolt patterns, and mechanically resistant. Taking in consideration these restrictions it is easy to understand why measuring the strain imposed on a selected piezo-resistive material, just by reading the voltage across the resistance of the material, is still the most used sensing technique. There are several ways and materials to design sensing gauges, being metal-wire, metal-foil and semiconductor gauges de most common. From those, the metal-foil gauges show some interesting features. With the developments in etching processes, metal-foil gauges became a very interesting possibility. They are manufactured in very thin foils (less than 10 ?m), with sizes down to 200 ?m, etched by photographic methods. Consequently, there are virtually no limits to the variety of possible geometries. This gives greater flexibility to design geometries, but also to the type o stressing at the surface of the elastic material component where the gauge will be attached. Metal-foil gauges are linear gauges, with very low transverse sensibility (less than 0.3%). Also, their thermal characteristics are better than the semiconductor and metal-wire counterparts. All these arguments explain why metal-foil gauges are ideal for force/torque sensing elements. Force/torque sensors manufactured by JR3 (the ones we use in this paper) use metal-foil gauges bounded to elastic rings as sensing elements, which explain their superior behavior. They are basically composed by:

1. The sensing part. It is composed by the sensing elements disposed in the 3 Cartesian directions (X, Y, Z), and the electronics necessary to read the raw values and transmit then to the host receiver board. If the electronics is inside the sensor, preferable situation for noisy or industrial environments, there is no analog signal being transmitted and high sampling rates can be achieved (8 Khz).
2. DSP receiver Board. Based on the same basic architecture several interfaces can be used. If the issue is high access rates, then fast IO buses must be used and a shared memory mechanism must be implemented to exchange data and program de sensor. JR3 offers several interface buses like VME, PCI (up to four channels per board), CPCI and ISA. The receiver boards are basically DSP boards that implement digital filters and dispose sensor information to users. Also they parameterize readings (offsets, full scales, geometrical transformations, etc) and implement a few interesting functions: maximum and minimum values (peaks), warning and error bits, etc.
   

   Fig. 1 — Force/Torque sensor overview (using PCI receiver board).
   
   
3. Interface software and drivers. For Win32 based operating systems we developed a complete set of tools that can be used to build applications using force/torque sensors. These tools include kernel drivers designed for Win32 operating systems, i.e., Windows 95/98/ME, NT4 and 2000. Basically, when we want to use some kind of equipment from a computer we need to write code and define data structures to handle all its functionality. We can then pack the software into libraries, which are not very easy to distribute being language dependant, or build a software control using one of the several standard languages available. Having in mind that force/torque sensors can be used by persons with different programming capabilities, and from several types of programming languages and environments, the collection of functions that access the sensor capabilities are offered in several packages: C++ Library, ActiveX software component, Matlab toolbox and Lab-Windows Virtual Instruments.

References

1. B. Siciliano, L. Villani, Robot Force Control, Kluwer Academic Publishers International Series in Engineering and Computer Science, Boston, MA, 1999
2. JN. Pires e JMG. Sa da Costa, "A Real Time System for Position/Force Control of Mechanical Manipulators", Proceedings of the 7th International Machine Design Conference, Ankara, Turkey, 1996.
3. J. De Schutter e H. Van Brussel, "Compliant Robot Motion I. A Formalism for Specifying Compliant Motion Tasks", The International Journal of Robotics Research, Agosto de 1988.
4. J. De Schutter e H. Van Brussel, "Compliant Robot Motion II. A Control Approach Based on External Control Loops", The International Journal of Robotics Research, Agosto de 1988.