Knowing some force-torque sensor system basics could boost your metallic friend’s performance.
Sensors and transducers measure and analyze changes off a system norm. Force and torque transducers generate output signals from vertical, lateral, and longitudinal forces and camber, steer, and torque moments. That’s handy on robots that rely on such information to successfully manipulate the surrounding environment. Also called F/T sensors, these devices sense up to six degrees of freedom and are critical to electrical and mechanical assembly, product testing, material handling, and other robotic applications. They verify part insertion, hold constant force during buffing, polishing, and deburring, and collect force information for lot testing and statistical process control. Utilizing strain gages or optical sensors, the sensor controller collects transducer strain gage vectors, performs computations, and outputs F/T data directly to the robot.
F/T Sensor selection
Many sensor systems for robotic applications are available, so selecting the right one is sometimes a challenge. Before choosing it’s important to fully define the application. How will the robot be used? What range of force will it experience? What is the environmental condition of the application — laboratory, assembly line, or someplace else? Once the application is clearly identified, components for the sensor system can be specified. Following a few steps and comparing transducer specifications to those of the application ensures optimized performance.
To select a transducer, first calculate expected moment and forces. Attached to the transducer is an end-effector that generates forces as it performs its task. Coupled with this force, the distance of the applied force from the transducer results in a moment. (Moment is the applied force multiplied by the distance from the transducer origin to the point at which the force is applied.) It is important to consider both overloaded and normal operating forces and moments affecting the transducer. In fact, moment capacity is usually the determining factor when choosing the best transducer for an application.
When calculating load on the transducer, be sure to include all loads the transducer will experience, including those loads the application does not monitor. Be aware that the published payloads of robots are typically the maximum loads for a published positional resolution. Because they’re typically overpowered, robots can actually handle and create forces many times greater than their load rating. During a crash, the inertia of the sudden deceleration can generate large loads, not to mention the force of impact. Typical robots can handle these conditions, generating 5 gs of deceleration during emergency stops. Handling greater forces does come with some loss of positional repeatability.
Strain gage sensing technology can also influence the transducer’s factor of safety. Transducers using high-output strain gages are designed to withstand higher overload conditions than those of lower output strain gages. High-output strain gages can also have lower noise levels, since they require less signal amplification. For these results, silicon strain-gages provide signals 75 times stronger than conventional foil gages.
Identify transducer capacity. Minimum and maximum force and torque, weight, diameter, and height must be known to select the correct transducer model and calibration. Typically sensor manufacturers provide selection tables that cross-reference measurement ranges with available transducer types.
Verify resolution and accuracy. Fine resolution requirements can conflict with moment capacity requirements. Transducers with larger ranges have coarser resolutions. A transducer’s output resolution is much finer than its absolute accuracy, so be sure the absolute accuracy fits the application. Like single- axis load cells, the absolute accuracy of six-DOF transducers is expressed as a percentage of rated full-scale load for each axis.
Sensor controller selection
Sensor controllers receive information from the transducer and produce resolved force and torque data. Onboard software then calculates the output data by multiplying the strain-gage vector by a calibration matrix to form the F/T data — consisting of three orthogonal forces and torque. After this step, force and torque data can be transmitted to the robot and serve as instructional signals to help the robot perform its intended function.
Continue on page 2
Most commercially available sensor controllers provide the following functionality:
• Output of all six axes of load data.
• Tool transformation. This allows movement of the center of origin to a user-specified location.
• Peak analysis. This detects and stores minimum and maximum F/T values.
• Biasing. This is a convenient way to subtract unwanted loads from readings.
• Data filtering. With this, users can minimize the effects of unwanted system vibrations.
• Programmable threshold monitoring. With optically isolated I/O connections, this provides highspeed response to discrete robot I/O panels.
Sensor controllers are selected according to sensor controller output resolution and format, as well as available software to interface with it. Some sensor controllers provide better resolution and noise performance than others: Commonly available output formats are RS-232 serial, analog voltage, computerbus ISA, PCI, and others. Manufacturers often offer interface software for easy system integration.
Basically, two fundamental types of sensor controllers are available: Stand-alone and computer bus. The advantage of stand-alone sensor controllers is that they’re selfpowered and self-contained, communicating with robot controllers via RS-232 serial formats or analog voltages. Their discrete I/O connections allow for easy connection to PLCs and other industrial equipment.
A computer bus sensor controller is targeted to a specific type of computer backplane, and plugs into robot and computer motherboards. Communication is through software drivers (such as the ActiveX for Windows platform) or directly to I/O mapped registers. Since computer bus sensor controllers can be placed inside robots they have a much cleaner appearance than stand-alone types. Often software provided by sensor controller manufacturers can display F/T information for all six degrees of freedom simultaneously on a computer screen, allowing users to easily modify different measurement parameters and determine the current loading.
F/T Sensor data use
The type of sensor controller selected is often dependent on how the F/T information will be used. It can be utilized in several ways: data collection and analysis, real-time force control, and threshold detection. A quick review of data usage types and corresponding sensor controllers helps finalize selections.
1. For data collection, computer bus sensor controllers provide easy integration for PC users. Installed in the PC, they communicate directly with standard operating applications such as LabView and Visual Basic. However, data collection speeds can be influenced by computer speed and the Windows operating system.
2. For real-time force control, ISA bus sensors are integrated with software drivers. All F/T data is available on the computer bus, allowing control software instant access. If users are not working in a PC environment, analog outputs created by stand-alone sensor controllers can be interfaced to any analog input card.
3. Force and torque threshold or limit detection is available on some sensor controllers. This capability allows the sensor controller itself to monitor transducer loads for specific loading conditions and notify the robot controller when the conditions have been met. Moving this monitoring function to the sensor controller relieves the robot controller of monitoring duties. (One use is to have the sensor controller monitor for dangerous loads.) When a dangerous load condition is detected, the sensor controller’s discrete output triggers the robot’s emergency stop circuit.
Once selected, a transducer must be electrically connected to its sensor controller. Usually manufacturers provide a standard length cable assembly. From there it must be determined how much more cable the application requires. Keep in mind that cables should be long enough to reach from sensor controllers to transducers in any robot position. If in doubt, longer cables should be specified to avoid cable breaking and possible system damage.
Continue on page 3
As previously stated, F/T sensor systems are used in many applications. Most suppliers manufacture rugged and extremely durable transducers, virtually eliminating environmental concerns from the selection process. Even so, physical attachments, mounting plates, and tool transformations are all factors to be considered.
Transducers are mounted to robots in several ways, with quick disconnects and interface plates. Other transducer options include specified temperature operating ranges, multiple calibrations, and unique operating environmental ratings (sometimes with proven resistance to MRIs and even nuclear radiation.) Working closely with sensor manufacturers optimizes selection. Sometimes only custom designed and built sensors meet specific applications.
Another integration type
Conventional sensor systems include a transducer, cable, and controller to convert strain-gage input to force and torque values. From there, output is sent to a computer or robot controller using serial, analog, or customized bus controllers. A few challenges to this setup:
• Serial communications are typically slow.
• Analog output must be converted to digital.
• Customized bus controllers are fixed to one bus format so changes in resolution, data speed, or format require expensive upgrades.
To address these problems, a new system from ATI integrates sensor systems directly into desktop and laptop computers. Transducer strain-gage signals are modified and sent to an off-theshelf data acquisition (DAQ) card that controls data speed and resolution. An interface card near the transducer produces a signal that can be read by most analog input DAQ cards, including USB , PCMCIA, ISA, PCI, and cPCI cards. The system can actually accept new hardware and software to match new bus formats and operating systems. Through the DAQ card the computer powers the interface card and transducer. PCbased software converts signals to force and torque output.
Software includes a Windows ActiveX component that configures the transducer system and converts raw voltage into force and torque. Non-Windows target platforms include a C library to perform necessary F/T conversions. The system can be used in most platforms that support ActiveX or Automation containment.
Foil versus silicon strain gages
The two most common types of strain measurement are foil and silicon strain gages. The electrical resistance of both changes in response to the amount of strain applied to them. Much like a rubber band that is stretched, gages get longer and thinner when they are pulled by tensile strain. The opposite is true when they are compressed; as they get thinner the reduced cross-sectional area increases electrical resistance.
Silicon strain gages also respond to this geometry change with a piezoresistive effect that is many times stronger than this first effect. The end result is that silicon gages can be much more sensitive to strain than foil gages. The measure of strain gage sensitivity is called gage factor. Typical foil strain gages have a gage factor of around two. In silicon strain gages with higher gage factors this value can reach 155. This means that devices that use the silicon strain gages can have much lower strains and consequently, much higher factors of safety. Despite these advantages of silicon strain gages, foil strain gages are more popular because they are less expensive than silicon strain gages and are easier to install.
Example: Data collection
Some automotive designers use systems on industrial robots to test automobile seats. For these setups, testing robots have a “fanny” form that is moved in and out of seats being tested. During this life-cycle testing, F/T sensors report data on the forces and moments encountered during each cycle recorded. This information is stored on the user’s system for analysis in case abnormal wear is noticed in the seat being tested.
Example: Real-time force control
Some machinists polish surfaces with robots that maneuver polishing heads along surfaces. In this application, the robot applies a constant force to the surface based on the data being reported by an F/T sensor. If the F/T system reports a light force, the robot will move into the surface to increase it; with a heavy force, the robot moves away slightly. In this manner a constant force is applied in real time.
Example: Force and torque threshold detection
Some users design automotive components for motion only in certain directions. For this objective robots move a test component through each of its positions. An F/T sensor measures the component’s force during this movement; good parts exceed set force and torque limits during motion, while bad parts either far exceed these values or offer almost no resistance at all. Good and bad levels are programmed into the F/T controller, so when the levels are met the controller outputs a digital signal indicating that. These signals are then fed back into the robot. This way the robot does not have to read actual levels, but only respond to digital signals.