Flight principles: how a multirotor holds itself still in the air

A propeller pushes air downwards and receives an upward thrust in reaction. Thrust scales roughly with the square of rotor speed and with air density, and a larger diameter produces it more efficiently (T = C_T × ρ × n² × D⁴, where ρ is air density, n rotor speed and D diameter). At high altitude or on hot days the air is thinner, so the rotors must spin faster to carry the same weight.

Hover is the state in which the total thrust of four (or more) rotors exactly balances the weight mg — on a quadcopter, mg ÷ 4 per rotor. The heavier the inspection camera or contact sensor, the faster the rotors spin and the less reserve thrust remains for climbing or fighting gusts. Inspection aircraft design is a battle over that margin.

A multirotor has no control surfaces. Every motion comes from differences in speed between motors. Slow the front motors slightly and speed up the rear, and the aircraft pitches forward; thrust tilts forward and the aircraft accelerates (T·sinθ in the figure). Roll works the same way. To stop, it tilts the other way and brakes.

Yaw works differently. A spinning propeller produces a reaction torque that tries to spin the airframe the opposite way. On a quadcopter, one diagonal pair spins clockwise and the other counter-clockwise so the torques cancel. Speed up one pair and slow the other while keeping total thrust constant, and only the torque imbalance remains — the aircraft rotates on the spot.

A multirotor is an inherently unstable vehicle that would tip over within seconds if left alone. The flight controller keeps it up with cascaded control. From the outside in, it sets targets for position, then velocity, then attitude, then body rate; the innermost rate loop reads the gyros and commands the motors. Most stages are PID controllers (correcting for the size of the error, P; its accumulation, I; and how fast it is changing, D) or simpler P-only stages (in PX4, for example, position uses P while velocity and rate use PID).

The trick is that inner loops run faster. Attitude errors grow in an instant, so the rate loop runs hundreds of times a second or more, while the position loop can be far slower. The mixer redistributes “rotate this much about this axis” into individual motor speeds. Poor tuning shows up near structures: the aircraft either over-reacts to small disturbances or responds too late and drifts.

Control works only if the aircraft knows its current state. No single sensor can be trusted on its own, so sensors with different characters are blended — typically with an extended Kalman filter (EKF) — to estimate position, velocity, attitude and even sensor biases at the same time.

The EKF splits the work: predict quickly with the IMU, correct slowly with everything else. What matters is whether it can detect a lying sensor and reject it. If GNSS jumps from multipath under a bridge, or the compass is skewed by steel girders, and the filter keeps trusting it, the aircraft can drift straight into the structure. See GNSS-denied environments, magnetic disturbance and the next chapter on navigation.

In wind, a drone holds position by leaning into the wind. The stronger the wind, the deeper the tilt and the more thrust goes sideways — leaving less margin for climbing or turning and draining the battery faster. A manufacturer’s “maximum wind” rating is a guide to holding attitude in open air, not a promise that it can photograph close to a structure in turbulent flow.

Structures bend, separate and shed the wind into vortices. Behind piers, at building corners and downwind of turbine towers, gusts change direction abruptly. The aircraft has to measure the drift before it can correct, so there is always a lag — and the distance it drifts during that lag sets the minimum stand-off from the structure. See wind and turbulence.

The same wind looks different from the side of carrying cargo. Compare it with how hovering and wind eat into delivery range (Airffic Logistics).

When a floor, ceiling or wall is close to the rotors, the airflow changes and so does thrust. Inspection is all about getting close to structures, so this cannot be avoided.

A multirotor consumes power simply by staying aloft. Hover power grows faster than weight, so adding battery also adds the power needed to lift that battery — a built-in dilemma. Lithium batteries store far less energy per kilogram than fuel, and typical inspection multirotors fly for tens of minutes at most.

• Cold: low temperatures reduce available power and cause sudden voltage sag. Warm batteries before take-off. Cold & icing
• Wind and weight: strong wind, heavy payloads and frequent climbs drain the battery faster than the estimate.
• The way home: in GNSS-weak areas or deep inside structures, returning takes longer. Set reserve thresholds by what it takes to get back.
• Battery health: capacity fades with cycles and storage. Keep a log per battery.

If you want numbers on how long a given payload and battery will keep you aloft before choosing an inspection aircraft, you can estimate hover time from airframe mass and battery (Airffic Architect). The results come from a coefficient model, not a thrust-stand measurement, so keep the two apart.

A fixed-wing aircraft generates lift from forward speed (lift = ½ × air density × speed² × wing area × lift coefficient). It does not need rotors spinning just to stay up, so it flies much further on the same energy — ideal for linear assets tens of kilometres long such as pipelines, power lines, railways and levees. But it cannot stop and get close.

Flying long lines almost always means beyond visual line of sight (BVLOS) operations. The barrier is usually airspace coordination and operating procedures rather than the aircraft itself — continued in remote and BVLOS operations.