CHANNELS

Aerospace

Automotive

Automated & Autonomous Vehicles

Battery & Electrification Technology

Defense

Drone TV

Electronics

Manufacturing & Prototyping

Materials

Medical

Motion Control

RF & Microwave Technology

Robotics & Automation

Sensors & IoT

Test & Measurement

Aerospace
Automotive
Automated & Autonomous Vehicles
Battery & Electrification Technology
Defense
Drone TV
Electronics
Manufacturing & Prototyping
Materials
Medical
Motion Control
RF & Microwave Technology
Robotics & Automation
Sensors & IoT
Test & Measurement
Aerospace & Defense
Search Results

Anti-Cogging Software Approach Makes Motors Smoother for Robotics

Smooth motion is critical to robotic applications like haptics or those requiring high-precision force control. These systems are often direct-drive, so any torque ripple in the motor output must be minimal. Unfortunately, low inherent torque ripple motors are expensive. Researchers from the Modular Robotics Laboratory at the University of Pennsylvania have come up with a software method to map and suppress torque ripple from cogging torque, so that low-cost motors can perform as well as expensive ones.

Topics:
Automation Motion Control Motors & Drives Robotics Sensors Software

More From SAE Media Group

    Transcript

    00:00:01 cogging torque Ripple minimization via position-based characterization or what we call anti- cogging lots of robots require high torque direct drive or low gear ratio actuators such as human safe back drivable robotic arms haptic devices and High Fidelity position control but the motors in these robots tend to be expensive why use motors like

    00:00:26 this when there are Motors that are rated to the same torque half the weight and when third the cost this motor should run smoothly but it doesn't because of torque Ripple across a wide range of Motors torque Ripple and cost are inversely related here on the x-axis we have cost on the Y AIS is the torque Ripple ratio which is a measure of torque Ripple and you can see cheaper

    00:00:50 motors have more torque Ripple in these applications computation and electronics often hold a large portion of the cost but with Mo's law and consumer electronics continuously driving down prices actuators become the dominant cost relatively speaking let's look at the three main sources of torque Ripple to see if we can help this problem mutual and reluctance torque are when

    00:01:11 current flows through the coils causing a magnetic field that reacts with the magnetic fields in the rotor ideally the shape of these currents with respect to motor position should be sinusoidal or trapezoidal to make the motor spin but in reality they aren't and this shape discrepancy causes a torque Ripple voltage and current sensors can measure this discrepancy cogging torque also

    00:01:30 called detent torque is when the permanent magnets attract the motor's core unlike mutual and reluctance torqus cogging torque cannot make the motor spin continuously only contributes torque Ripple is always present and is only detectable by mechanical sensors torque Ripple has been heavily studied in the last 25 years and most of the research falls under two categories

    00:01:50 software methods for Mutual and reluctance torque Ripple minimization and Hardware redesign solutions for cogging torque Ripple minimization but can we combine the two philosophies to minimize cogging torque Ripple with software Holton spring laid down the foundation for Cog mapping by iteratively monitoring current while a speed Loop tried to hold speed constant

    00:02:12 we extend this to two methods without requiring a speed controller current sensor or iteration the first method is to use position control at all positions and map the applied voltage or current when the motor is stopped we can also compensate for static friction or stion and transistor Deadtime with this method the second method is to spin open loop

    00:02:34 as slowly as possible and map the accelerations you can see these two maps agree nicely with each other but to double check we used a force torque sensor during testing to validate the two methods now we can invert the Cog map and feed forward these torqus currents or voltages we tested both methods on 11 Motors this is the same graph as before

    00:02:56 with the price versus torque Ripple ratio but now with the anti- Cog Motors you can see what was once an inverse relationship is now roughly constant with an average of 69% reduction we think pwm resolution is the cause of this constant you can see Peak to- Peak torque Ripple is on the same order of magnitude as the pwm resolution between 1 and 5 1/2 counts so we doubled the

    00:03:19 resolution a motor that had 69% reduction went to 82% reduction simply by doubling the resolution to verify that anti cogging Works in robotic applic ations we made a robotic arm modeled after a phantom Omni we looked at the motors with a torque rating within a factor of two of the Phantom omni's continuous torque then rank them by performance and economic

    00:03:40 value which we defined as the inverse of peak-to peak torque Ripple times cost the naturally smooth eflight Park 400 has the best value but after anti- cogging the naturally rough Exceed RC Rocket 400 has about the same peak-to Peak torque but only onethird the cost and was the cheapest tested motor we used an updated motor Drive driver with an even higher voltage resolution as

    00:04:02 well as wireless communication so that no wires cause friction by crossing over joints we ran the arm along various trajectories including a 36 line segment 2 meter trajectory where the arm had a root mean squared error of 7.4 mm without compensation while with compensation the error was 3.5 mm a 52% reduction you can see the nominal Motors get caught on cogs particularly in the

    00:04:28 top right of the m and at the bottom of the o while the anti- Cog Motors Traverse these regions smoothly we have also applied anti- cogging to permanent magnet and hybrid type steer motors with no changes to the algorithm so here's the anti- cogging turned off and here's anti- cogging turned on and here is the desired trajectory along with the same two paths but gathered

    00:04:50 from encoder values as you can see you can now get good performance from cheap Motors thanks to you our funding sources and everyone that worked on this project