Robots do not move perfectly on the first try because motors, gears, loads, and surfaces all introduce small errors. Feedback control is the method robots use to notice the difference between where they are and where they should be. In a robot arm, this can mean reaching past a target spot, measuring the mistake, and moving back until the gripper is aligned.
This idea matters in factories, drones, self-driving cars, and medical robots because accurate correction makes machines safer and more useful.
A feedback control loop starts with a desired position and compares it with the actual position measured by a sensor. The difference is called error, and the controller uses that error to decide how strongly and in which direction the motor should move. If the arm overshoots the target, the sign of the error changes, so the controller commands a correction in the opposite direction.
Good feedback control reduces error quickly without causing endless wobbling or unstable motion.
Understanding Feedback Control
A controller does not make one final decision and then stop. It repeats a cycle many times each second. The speed of this cycle is called the update rate.
A fast update rate can help a robot respond before an error grows large. However, fast updates do not fix a poor sensor. Measurements can contain noise, which is small random variation in the reported value.
If the controller reacts strongly to every tiny change, a motor may twitch or vibrate. Engineers often smooth sensor readings or ignore very small errors near the target. This small allowed region is useful when perfect stillness is not realistic.
The controller setting called gain determines how much effort is used for a given mistake. With low gain, motion is gentle but slow. A robot may take too long to reach its goal.
With high gain, motion begins strongly, yet momentum can carry the mechanism past the goal. The correction then reverses, which can create repeated swinging. More advanced controllers use three kinds of action.
Proportional action responds to the current error. Integral action adds up error over time, helping overcome effects such as friction or a constant load.
Derivative action responds to how quickly the error is changing, helping slow the system before it overshoots. Integral action must be limited because a motor that is already at maximum effort cannot obey extra accumulated commands.
Real machines have limits that a simple control rule must respect. A motor cannot produce unlimited force, and a battery voltage can fall during use. Gears introduce backlash, which is a small free movement before gear teeth press against each other.
A wheeled robot can slip on dust, carpet, or a slope. A drone changes behavior when it carries a camera or meets a gust of wind.
Feedback helps handle these changing conditions because it works from measured results rather than assuming that every command produces the same motion. Similar ideas appear in a 3D printer keeping its nozzle on a path, a phone stabilizing its screen, and a car maintaining a selected speed.
When studying a control loop, follow the direction and units of every signal. A position target measured in degrees must be compared with a position reading in degrees. A reversed sensor direction creates a dangerous problem because the controller can push farther away from the goal.
Graphs are especially useful. Plot the target, actual value, error, and motor command against time. Look for slow response, overshoot, steady offset, or noisy motion.
Change one setting at a time during tests. Start with low motor power and keep clear of moving parts. Control design is usually a process of measuring, adjusting, and testing again under realistic loads.
Key Facts
- Error = target position - actual position
- Control signal = controller output sent to the motor or actuator
- Negative feedback means the correction acts to reduce the error.
- If error = 0, the robot is at the desired position and no correction is needed.
- Proportional control can be written as u = Kp e, where u is control signal, Kp is gain, and e is error.
- Sensors close the loop by measuring actual position and sending that information back to the controller.
Vocabulary
- Feedback control
- A method where a system measures its output and uses that measurement to correct future action.
- Setpoint
- The desired value a system is trying to reach, such as a target position for a robot arm.
- Error signal
- The difference between the desired value and the measured actual value.
- Controller
- The part of a system that calculates what action to take based on the error signal.
- Sensor
- A device that measures a physical quantity, such as position, speed, angle, or force.
Common Mistakes to Avoid
- Thinking the robot knows it made a mistake without a sensor is wrong because feedback requires a measurement of the actual position.
- Using error = actual - target without checking the sign is wrong because the controller may command motion in the opposite direction from what was intended.
- Making the controller gain too large is wrong because a strong correction can cause overshoot, oscillation, or instability.
- Confusing open-loop control with feedback control is wrong because open-loop control sends commands without checking whether the robot reached the target.
Practice Questions
- 1 A robot arm has a target position of 50 cm and its sensor reads 57 cm. Calculate the error using error = target position - actual position. State whether the arm overshot the target.
- 2 A proportional controller uses u = Kp e with Kp = 4. If the error is -3 cm, what is the control signal u? What does the negative sign mean for the motor direction?
- 3 A robot arm keeps moving back and forth across a target instead of settling on it. Explain what might be wrong with the feedback control loop and name one change that could improve its motion.