🦾 The robot that "feels": Why adaptive control is a revolution in manufacturing

Introduction

Until recently, the robot was essentially "blind" and "deaf". It could repeat the same movement thousands of times with incredible accuracy. The problem arose as soon as the part was a millimeter off, or the surface was not perfectly flat. The robot "crashed" or made a mess.

This "dumb" automation required extremely expensive and precise fixtures to ensure that each piece was always in the same place.

But what if a robot could "feel" the surface and apply just the right amount of pressure? What if it could "see" a weld seam and follow it itself, even if it's not perfectly straight? That's exactly what adaptive control is - the ability of a robot to respond to the real world in real time.

⚙️ Problem: A perfect robot in an imperfect world

For every technologist, variability (inaccuracy) is the enemy. And the real world is full of variation.

What it looks like in practice: You want to automate the grinding of a weld on a casting. You program the robot to follow a precise path. But the first casting is 0.5 mm larger than the second.

The most common problems and losses:

• ➡️ "Stupid" robot: On the first piece, it grinds too much (waste), on the second, it doesn't even touch the surface.
• ➡️ Extreme costs: To prevent this, you have to invest hundreds of thousands in fixtures that clamp the part exactly to the hundredth of a millimeter every time.
• ➡️ Impossibility of automation: Tasks requiring "feel" (grinding, polishing, fine assembly) were considered unautomable. They remained with humans.

🤖 How adaptive steering helps (Touch and Vision)

Adaptive control gives the robot senses. It turns it from a "repeater" to a "craftsman". There are two main types:

1. Force Control (Hapticity) The most commonly used force and torque sensor (Force/Torque Sensor), which is located on the robot's wrist.

   How it works: You no longer program exact coordinates. You program the resulting force. You tell the robot, "Hold the sanding belt on the surface with a constant force of 5 Newtons."

   Result: It doesn't matter if the surface is curved or offset. The robot adapts itself – it goes deeper or retracts to always maintain exactly 5 N.

2. Image control (Vision) The robot is equipped with a camera (2D or 3D) that scans the work surface.

   How it works: You want to weld a joint between two sheets of metal. The camera "looks" directly at the weld seam just in front of the torch.

   Result: Even though the sheets are warped by heat, the camera sends corrections to the robot in real time ("move 0.2 mm to the left, now 0.1 mm to the right") and the robot follows the real gap itself.

📈 Key benefits: Mainly quality and savings

• 1️⃣ Automating "human" tasks
  Finally, you can effectively automate processes that used to require human touch: grinding, polishing, deburring, fine assembly (inserting connectors "by feel").
• 2️⃣ Perfect and consistent quality
  Every part is ground with the same force. Every weld is perfectly centered in the joint. Eliminate quality variations caused by operator fatigue.
• 3️⃣ Saving on product costs
  You no longer need extremely expensive and complex fixture fixtures. The robot "finds" the part with a camera or "touches" it with a force sensor.
• 4️⃣ Higher security and speed
  During fine assembly, the robot "feels" that the part does not fit and does not force it in (it does not destroy it). During grinding, it maintains optimal pressure for the highest material removal rate.

🧠 What does a real deployment look like (Typical scenario)

Before: A stainless steel tank manufacturing company. After welding, operators had to manually grind and polish the welds to a mirror-like shine. It was an incredible amount of work, in the dust, and with inconsistent results.

After deployment (Adaptive control):

• ✅ They mounted a grinder on the UR10e robot (which already has a force sensor in it).
• ✅ The operator "taught" the robot the approximate weld path.
• ✅ The robot then "touched" the weld itself and started grinding with constant pressure.
• ✅ Result: The process is 3x faster. The resulting surface is perfect every time. And the operator focuses on checking instead of hard work.

📦 Recommended technologies for adaptive robotics

• UR10e – Universal Robots collaborative robot It is the perfect example of a force-controlled robot. The entire "e-Series" has an integrated force and torque sensor right in the wrist. It is therefore "sensitive" by nature and ready for grinding or fine assembly without the need to buy expensive external sensors.
• Dobot CR10 - flexible robotic arm Dobot is a great platform for adaptive vision control. Thanks to its high accuracy and open API, it is very easy to connect it to any 2D or 3D camera system for tasks such as path following or bin picking.
• Force sensors (e.g. OnRobot HEX): Even if your robot (like Dobot) doesn't have a force sensor built in, you can add one. External sensors (e.g. from OnRobot) attach to the wrist and give the robot the "feel" it needs for grinding.

❓ Frequently Asked Questions (FAQ)

Is programming a robot with a force sensor complicated? Paradoxically, it is often easier. Instead of programming 1000 points on a complex curve, you just "guide" the robot with your hand and say in the program "Hold a force of 5N in this direction". The robot will adjust the path itself.

What if my part is in a completely different place every time? If the part is in a different place every time (e.g. chaotically in a container), force control is not enough. Here you need image control (3D camera) that tells the robot where the part is. Adaptive control (force) only comes into play after it has grasped the part.

Is it worth it? Sensors are expensive. Sensors are expensive. But so are scrap and downtime. And the most expensive are special fixtures. The investment in a sensor often pays for itself within a year just by saving on one complex fixture that you didn't have to make.

🧭 Conclusion

Adaptive control is what turns a “dumb” robot into a smart colleague. It’s the technology that allows you to automate tasks you thought only a human could do. Whether you need a robot to “feel” the pressure of a polishing job or “see” a weld seam, adaptive control is the path to higher quality and huge cost savings.

Find out how robotization can help your company - visit svet-robotu.cz and discover solutions for smart and responsive automation.