Try it / design a sensor package
Choose the measurements.
In clear, well-lit water, a pilot needs to read a painted label while monitoring the vehicle’s depth.
Choose a mission and sensors, then check your reasoning.
A simplified design exercise, not an equipment specification or a complete navigation system.
Measurement comes before a decision
A camera gives a view; it does not automatically understand the scene. A pressure reading can help estimate depth; it cannot tell you whether you have found the right object. Interpretation, calibration and control complete the path from a signal to a useful action.
- MeasureA sensor produces a reading.
- InterpretA person or program checks what it means.
- Act & checkThe system responds and reads again.
Five sensor types and their limits
| Sensor | Useful information | A limitation |
|---|---|---|
| Camera | Visible details, colour and markings | Needs usable light and visibility; an image alone does not provide a calibrated depth reading. |
| Pressure sensor | Water pressure, used to estimate depth | Needs a surface reference and calibration; it cannot identify a painted label. |
| Imaging / ranging sonar | Acoustic returns used to estimate range or shape | Resolution, geometry and reflections matter; it will not tell you the colour of a label. |
| Inertial measurement unit | Acceleration and rotation, used to estimate motion and orientation | Estimates can drift; an IMU alone is not an absolute position or depth measurement. |
| Temperature sensor | Temperature at the sensor | It cannot show an obstacle, identify a label or measure depth on its own. |
On land, robotics projects also use switches, wheel encoders and optical or ultrasonic distance sensors. Choose hardware for the actual environment: a classroom air-distance sensor is not interchangeable with underwater sonar.
What did Stinky use?
The Carl Hayden team’s 2004 report describes cameras and a tether that carried video, audio and control signals. Operators used feedback at the surface to guide the vehicle. See the documented Stinky design for its motors, onboard battery and control loop.
The five options in the activity are a modern design exercise. They are not a claim that the original team fitted all five sensors. Keeping a historical design separate from a proposed design is part of using evidence well.
More sensors do not automatically make a better robot
Every added device needs power, mounting, a connection and a way to interpret its output. It can also fail. Two sensors are helpful when they answer different questions or provide an independent check; adding devices without a purpose can make a system harder to understand.
- Name the measurement. Write “distance to the wall,” rather than “awareness.”
- Identify the conditions. Will water clarity, lighting, motion or reflections affect it?
- Plan a check. Compare a reading with a known reference before trusting it.
- Describe a missing reading. Decide what the controller should do when data is unavailable.
A 15-minute classroom challenge
Choose one mission above. Spend five minutes selecting sensors, five explaining a limitation of each, and five changing the conditions. If clear water becomes murky, which decision would you revisit? If the temperature sensor responds slowly, what must change about your measurement procedure?
There can be more than one workable design. Assess whether the student connects a sensor to a needed measurement and explains uncertainty. Continue with the missing-reading experiment or compare human and model decisions in robotics vs AI.
Sources & further reading
- University of Rochester: robot construction and sensors — sensing principles and device examples.
- NOAA: underwater robots — vehicles, sensors and exploration resources.
- Carl Hayden’s original technical report — the evidence for the historical vehicle.
These original missions teach selection and reasoning. They do not specify a complete vehicle, calibration procedure or physical build.