A builder wiring a first robot grabs whichever sensor is nearest, then wonders why readings mislead. The main types of sensors split by what they measure and how they report it; pick by the job and mounting spot, not by reputation.
Sensor choice is one decision inside the wider task of how to build a robot. This guide sorts the main sensor families, clears up common misunderstandings and turns the differences into selection rules for robot builds.
Sorting sensors by method and output
There is no single master list of sensors, because several classification schemes run side by side. Yokogawa classifies sensors by measurement method into active and passive types. Passive sensors need no energy source of their own to detect a measurand, while active sensors require an external energy source. A second scheme, described by ElProCus, divides all sensors into analog and digital.
A third scheme groups sensors by the quantity they sense. Yokogawa’s mechanical group covers pressure, flow rate, vibration, distance, velocity, acceleration and force. Its radiation group covers infrared, X-rays, visible light, ultrasonic, acoustic and radio waves, which is why infrared and ultrasonic sensors sit in the same family.
These schemes overlap rather than compete. A single part can be described by its measurand, by whether it needs its own energy and by whether it reports an analog or digital signal. Read a datasheet with all three questions in mind before comparing parts.
Comparison of the core robot sensors
| Sensor | What it senses | Planning use |
|---|---|---|
| Accelerometer | Vibration and shock; linear acceleration that feeds orientation estimates | Fit when the build must detect impacts or shaking, and pair it with a gyroscope when orientation matters. |
| Gyroscope | Rotational velocity, used alongside accelerometer data to estimate orientation | Choose it for turning motion; it is rarely useful as the only motion sensor on a build. |
| Encoder | A digital sensor used for discrete counting, linear and rotary encoding, and position measurement | Mount it where the robot needs to know how far a joint, wheel or slide has moved. |
| Camera | Single and continuous 2D images | Choose it when the robot needs image data rather than one distance reading. |
Descriptions of accelerometers, encoders and cameras follow Dewesoft’s sensor overview, and the orientation and rotation details follow the Nimble Physics notes on accelerometers and gyroscopes.
Motion and position sensing
The first misunderstanding is that an accelerometer tells a robot where it is. According to Dewesoft, accelerometers are used for measuring vibration and shock. The Nimble Physics documentation explains that accelerometers and gyroscopes together are used to estimate the orientation of a device, working from sensed linear acceleration and rotational velocity. Orientation is something the system computes from these readings; neither sensor reports it alone.
Encoders answer a different question. They fall under the digital sensors that Dewesoft lists for counting, linear and rotary encoding, and position measurement. A rotary encoder suits a wheel or a rotating joint, and a linear encoder suits a sliding axis. Inertial sensors describe how the whole body moves, whereas an encoder describes how far one mechanism has moved. Many robots need both kinds, because the two answers do not substitute for each other.
The practical rule is to decide whether the question concerns the body or a mechanism. If the robot needs to know its tilt or turning, start with the inertial pair. If it needs to know how far a motor shaft has turned, use an encoder.
Distance and vision sensing
Ultrasonic sensors measure distance from echo timing. As ElProCus explains, the time taken to transmit and receive the echo is used to determine the distance to an object. The output is one distance value, which is simple for a controller to act on.
Infrared sensors fill a similar role. The same source notes that IR sensors can be used for detecting obstacles of a robotic vehicle. Both ultrasonic and infrared devices answer how far away an obstacle is, or whether one is present, without producing a picture.
A camera works differently. Dewesoft describes camera sensors as capturing single and continuous images, so a camera supplies image data rather than a direct distance measurement. That makes it the right choice when the robot has to recognise or follow what it sees. It is the wrong choice when a single reading is enough, because the images must still be interpreted before the robot can act on them.
Choosing sensors for a robot build
Yokogawa states that sensors are selected according to intended use and installation location, and that the differences between sensor types do not make one superior or inferior. That statement is the foundation for the rules below.
Misunderstandings that lead to poor choices
- “One sensor type is best.” No type wins in general. A gyroscope cannot measure distance, and an ultrasonic sensor cannot measure rotation. Start from the quantity and the place it will be mounted.
- “Active is better than passive.” Active sensors need an external energy source, while passive sensors do not need their own. If a build budgets power tightly, count the energy an active sensor requires before committing to it.
- “A sensor reads the same anywhere.” The installation location is part of the selection. A part that suits a clean bench may be a poor match for the spot where the robot actually carries it.
- “Thermocouples suit every temperature job.” Dewesoft calls the thermocouple the most popular temperature sensor overall because of its relatively low cost and reliability. Popularity is not the same as fit, as the typical figures below show.
Temperature sensors as a worked comparison
Dewesoft’s typical figures show how much two sensors of the same kind can differ:
- Thermistor: typical measuring range -100 to 325 °C and typical accuracy 0.05 to 1.5 °C.
- RTD: typical range -200 to 650 °C and typical accuracy 0.1 to 1 °C.
- Thermocouple: typical range 200 to 1750 °C and typical accuracy 0.5 to 5 °C.
Read these as typical values rather than guarantees; the datasheet for the exact part governs. The thermocouple’s range is far higher but its typical accuracy band is wider, so hot zones point towards it and tight accuracy at moderate temperatures points towards a thermistor or RTD.
A selection sequence
- Name the measurand. Decide whether the robot must sense motion, position, distance, an image or a temperature.
- Check the datasheet limits. Compare the part’s range and accuracy with what the task needs.
- Count the energy. Establish whether the sensor is active or passive, and whether the build can supply what it needs.
- Match the output. Confirm the controller can read an analog or digital signal from the part.
- Test it in place. Judge the sensor at its real mounting location, not on a bench.
Sensor combinations on a robot
Robots use a mix of the sensors above, chosen job by job. The pairings below are build suggestions that follow from each sensor’s measurand rather than a fixed recipe:
- Drive feedback: rotary encoders on the wheels or joints report how far each mechanism has moved.
- Body attitude: an accelerometer and gyroscope combination supplies the data for orientation estimates.
- Obstacle awareness: an ultrasonic or infrared sensor flags objects ahead, which is enough for simple avoidance.
- Scene understanding: add a camera only when the task needs images, since it brings data that must be processed.
Start with the smallest set that answers the robot’s real questions, then add a sensor when a specific measurement is missing.
Next steps for builders and students
Sensor knowledge is a core skill for anyone moving into the field. The career route is laid out in How To Become A Robotics Engineer: A Practical Step-By-Step Career Guide For 2026. For the academic side, including where sensing and controls fit in a programme, see Robotics Engineering Degree: Your 2026 Roadmap To Jobs, Specializations, And Pay.
Frequently asked questions
| Question | Answer |
|---|---|
| What does an IMU contain? | An inertial measurement unit contains an accelerometer and a gyroscope, and often a magnetometer as well. The Nimble Physics documentation notes that IMUs are common in motion capture and in wearables such as smart watches. |
| Does a temperature sensor need external power? | An RTD must be powered by the measuring system, whereas a thermocouple is self-powered. Dewesoft adds that thermocouples give a very small microvolt or millivolt output, so electrical noise and interference can occur. |
| Which quantities count as mechanical measurands? | Yokogawa lists pressure, flow rate, vibration, distance, velocity, acceleration and force. |
| Can ultrasonic sensors work where access is difficult? | Yes. ElProCus describes ultrasonic sensors measuring the distance of an object in inaccessible or harsh areas. |
