Can a motor be used for haptic torque feedback in Surgical Robotics?

The quick answer is yes, but there are several things to consider. They include motor type and design, mechanical system design, and control system design. This article is an introduction to the advantages and challenges of motoring while sensing torque from the same device.

Using Motors in Haptic Systems for Torque/Force Feedback

With more things being remotely controlled, it has become important for the user to get torque or force feedback, (haptic feedback). One well known requirement is a result of fly by wire aircraft where all flight surfaces are remote. Another area is robotic surgery where the surgeon may be sitting far from the robot and the patient. Providing the correct amount of force or torque feedback is paramount allowing the surgeon to “feel” the objects and possible preprogram “keep out” regions.  

It is common to use mechanisms and electric motors to provide force feedback as well as soft travel limits. This type of haptic feedback greatly depends on the smoothness of torque production over the complete range of motion. The motion translation, (rotary to linear, for example), must be balanced and smooth. Motor cogging torque, friction, damping, and torque ripple in the motor itself will directly affect the usability “feel” of the system.

• Cogging torque is an unenergized cyclical torque disturbance as the motor shaft is rotated, (or in the linear case the motor moved linearly).  Cogging torque is present in most permanent magnet motors with iron laminations containing wound teeth. Motor designers take measures to minimize cogging torque by design, but these mitigations do not remove all the cogging especially under varying operating conditions. Cogging torque it is not routinely part of final inspection and can vary greatly based on tolerances and the build processes. It is typically not on a motor datasheet.

• Friction is torque independent of direction. It is typically associated with bearings and mechanical parts that are in contact. Electric motors also have an internal friction component from magnetic field induced hysteresis in the iron laminations. Hysteresis losses are proportional to the rate of change of magnetic fields in the steel and the flux levels. A motor with high-energy rare earth magnets will exhibit higher hysteretic losses, so when you are looking for high performance and high torque density, beware that high hysteresis comes with the package. Hysteresis loss is another parameter that is not measured on final test and cannot be found on most datasheets.

• Damping is a torque proportional to the movement, or speed. It is typically associated with viscous bearing losses. However, almost every motor has an internal damping component that is related to eddy currents in the iron laminations. Motors with iron are laminated specifically to reduce eddy currents. More thinner laminations increase the resistance to eddy current flow. The damping torque or force is proportional to speed or the frequency of the changing magnetic field.  Eddy current damping is rarely shown on a motor datasheet.

• Torque ripple is also present in every motor. Torque ripple is a result of a mismatch between motor phase related torque, (torque versus angle profiles), and driver current profiles. Most surgical robots use synchronous brushless permanent motors, (also called BLDC). These motors generally produce torque proportional to current and proportional to angle. In servo control system, the torque angle is fixed using the encoder and commutation algorithm, therefore, applying current produces torque. If the motor torque versus angle curve and the drive current versus angle curves are not perfectly matched torque ripple will be present.  Torque versus angle profiles are not measured in final test and they are not provided on motor datasheets.

• Torque linearity is a problem for most motors. This phenomenon is where torque output is not linear with current applied (assuming a fixed commutation angle). All motor datasheets provide a “Kt” or torque constant, implying that the torque output will increase at a linear, constant rate with current. In reality most motors have internal saturation, and torque is not proportional to current and can roll off 5-20% in the operating range. This phenomenon forces the torque profile to not match the current profile and causes torque ripple.  Torque linearity is not shown in motor datasheets.

Almost every point above has a statement indicating that the important information is not usually shared by the motor suppler. Interesting!

1. Slotless or Ironless motors are the best technology for torque sensing/torque feedback.

a. More specifically, slotless motors without any iron are the only technology that will allow for pure torque sensing. Air-core linear or Air-core linear-arc motors are the equivalent for force control and force feedback.

b. These technologies exhibit zero cogging torque and excellent torque linearity.

c. If designed properly they have sinusoidal torque versus angle curves, and sinusoidal force versus position curves (linear).

d. They exhibit zero eddy currents and hysteresis torque.

e. Haptic systems are human operated and therefore do not typically need high torque or force. This supports a direct drive approach, or a minimum ratio translation between rotary or linear motion.

f. As with all motors, calibrating to eliminate manufacturing variability in mechanical parts and magnets is necessary to reduce the +/- 10% parameter tolerance down to +/- 2%.

g. With selection of the slotless motor, total variability can be in the +/-2% after calibration.

2. What if a Slotless motor does not have enough torque in the size needed?  

There are many traditional torque motors with good characteristics for torque sensing. Unfortunately, you will have to engage with the supplier to find out details on the motor beyond the datasheet.

a. Cogging torque needs to be measured, you will need to know the amplitude and frequency of the cogging. The challenge becomes compensating for higher frequency harmonics and unit to unit variability. Cogging reduction algorithms can help if cogging is repeatable and consistent. It is common to develop an algorithm to reduce cogging which works well until the next shipment of motors arrives.

b. Torque linearity. The torque output within the range of operation needs to be linear with current. Most motors are designed to be within 5% torque linearity at thermally rated current and can fall off by 10-15% at higher levels of current. You will need data to make your decision.

c. Magnetic eddy current and hysteresis losses can greatly impact torque sensing ability. For example, if 10% of the current is going to overcome these loads, only 90% of the current is actually going into moving the robot.

d. Bottom line is that selecting a motor from the internet based on a datasheet is not advised if you desire to use motor current as an indication of torque. Adding up the unknowns might yield; 5% error for cogging torque, 5% for linearity, 5% for core losses (hysteresis and eddy currents), +/-10% for normal parameter variation. In total it could result in +/- 25% error before calibration, and 15-17% error after calibration. 

The best system is always an engineered solution. If you are looking to attain performance and torque or force sensing feedback, then it is important to acquire the right equipment to test your components. A low friction (near zero) torque transducer system, (or a force gage with appropriate sensitivity for linear), along with precision encoders, will allow you to map torque versus position, cogging torque, and torque ripple.  Back driving with the motor with a torque or force transducer will allow you to measure losses.  Of course, working with engineers that have experience with motor design, operation, and test, while covering things that are not on the datasheet, is always the best approach.

Figure 1 – Slotless Motor from ThinGap