Collaborative Robot

As human-robot collaboration becomes more common, expectations for robots have risen significantly. Traditional large industrial robots used for palletizing, material handling, and welding must operate behind safety barriers and cannot share a workbench with people.

Collaborative robots, by contrast, bring lightweight, safe, and convenient automation to light-industry tasks. They handle delicate objects more dexterously and are better suited to working alongside people. Meeting the standards of human-robot collaboration requires higher performance in precision, size, safety, and usability.

Actuator

Every robot joint relies on a motor, reducer, encoder, and drive. Traditional large robots have fewer space constraints, so these functional modules are often assembled as separate components to reduce cost.

Collaborative robots have much tighter space constraints. Their functional modules therefore need to be integrated into compact modular actuators installed at each joint. This arrangement produces a smaller overall structure, simplifies production and maintenance, and helps keep the system cost-effective.

As shown above, every moving joint contains an integrated actuator that executes commands from the central controller. In addition to providing sufficient torque, a collaborative robot places higher demands on the following capabilities:

• High precision: The robot's motion accuracy depends on the accuracy of every actuator. Even with high-quality mechanical parts, precise position feedback from the encoder is essential for accurate actuator control and coordinated robot motion.
• Safety: A collaborative robot must sense external force. Just as a person feeling their way with closed eyes stops when an arm encounters an obstacle, the robot must detect a collision and stop safely.
• Ease of use: Collaborative robots should not require the complex workflows or dedicated operators associated with conventional industrial robots. When a person guides the robot arm by hand, the robot must sense the direction of the applied force and follow it. This drag-teaching capability makes setup more intuitive.

Safety and ease of use both depend on sensing external forces acting on the robot body. At each joint, this becomes a requirement to sense external torque. Higher torque sensitivity allows the actuator to detect smaller external loads and provide better control feedback.

This article focuses on the actuator arrangement illustrated above. The following sections explain the solution in detail.

Torque-Sensing Solutions

Common torque-sensing approaches include:

• Motor current: External force is estimated from the motor winding current. This method is convenient but provides limited accuracy.
• Dual encoders: A high-resolution encoder measures changes in the output position so that torque can be calculated. This approach combines low cost with high accuracy.
• Torque sensor: A dedicated torque sensor is installed at the output. This approach is costly and requires more complex algorithms.
• Strain gauges: Multiple strain gauges are bonded to the harmonic drive flexspline to measure torque. The structure is highly complex and difficult to calibrate.

Challenges of a Dual-Encoder Layout

Adding a second encoder at the output may appear straightforward, but the conventional arrangement above introduces several design challenges:

• Position of the encoder and output flange: To measure changes at the actuator output, the encoder stator must be fixed to the actuator stator. This forces the output-side encoder into the space between the output flange and actuator stator, complicating the structure. A longer output shaft also increases bending moment and can reduce structural stability.
• Encoder wiring: Because the encoder is located on the rotor side of the actuator, its power and communication lines must pass through a slip ring to reach the stationary side. This increases the number of slip-ring circuits and the overall package size.

Dual-Encoder Solution

A more effective topology uses a hollow-shaft design to move the output-side encoder to the actuator's stationary side. A return shaft connected to the output passes back through the hollow shaft and supplies rotational data to the encoder. This solves both the encoder-stator mounting problem and the wiring problem.

In this structure, Encoder 1 provides motor-control feedback, while Encoder 2 measures changes in output position.

The design uses KingKong Technology PCB magnetic encoders to provide the required resolution and accuracy. Their ultra-thin construction also enables a very compact layout: together, the two encoders occupy only about 15 mm in the axial direction.

Dual-Encoder Algorithm

This type of actuator commonly uses a harmonic reducer with a reduction ratio ranging from several dozen to approximately 100:1.

• Static operation: When external torque is applied to the output, Encoder 2 detects a small displacement. Because of the high reduction ratio, Encoder 1 on the motor side may not yet register a change. Comparing the two signals reveals even a slight force applied to the robot arm. In drag-teaching mode, the drive commands the motor to follow the applied force. Otherwise, the motor compensates for the disturbance to hold the output position.
• Dynamic operation: Encoder 1 data is used to calculate the theoretical output-shaft position. Comparing that value with Encoder 2 data reveals the external load, which is then supplied to the drive for subsequent control.

During motion, current sensing, torque sensors, and strain gauges all infer torque from changes over time. A dual-encoder system instead uses a spatial reference. Because a robot arm has a long reach and its center of gravity and momentum continually change, time-based measurements fluctuate throughout motion. Identifying a small external-force component within that changing curve is less precise than comparing positions in space.

In practice, dual-encoder spatial measurement is often combined with time-based motor-current measurement to produce better output data.

Future Development

KingKong Technology continues to develop more compact solutions. The patent-pending topology shown above is designed to reduce the package height to approximately 10 mm while simplifying wiring and control. Contact us for the latest development information.

This case study has presented a dual-encoder actuator solution for collaborative robots. Please contact us if you have questions or need support with your application.