Autopilot concepts

Levels of Automation

Stabilization

• Main article at Gyro

Stabilization involves preventing the aircraft from rapidly turning upside down or going into a dive: providing some minor feedback to the control surfaces rather than simply holding them steady. This may involve accelerometers, or accelerometers and gyroscopes. The broad term for this is gyro, short for 'gyrostabilizer'. On a fixed-wing hand-launch glider a gyro is often used during launch. On powered planes, it may be used to make 3D flight and aerobatics easier or simply to reduce thumb fatigue from rapid corrections in long-distance flight. Helicopters commonly use a gyro for fast reaction time, and add derivative functionality like flybarless controls and heading hold. The most basic multirotor flight controllers tend to include gyro functionality at a bare minimum, because of the difficulty of controlling multirotors without high-frequency feedback loops.

• Using Gyros on an RC airplane RCgroups thread

Altitude Hold

Altitude hold involves enough pitch/throttle control to keep the aircraft at a given altitude. GPS has a very rough altitude signal (particularly outside of the WAAS-covered US), and it is difficult to judge glide angles based on a gyro pitch in varying winds, so this usually involves a barometer.

Heading Hold

Heading hold adds absolute yaw stability (in relation to true north), and requires a magnetometer.

Control Axis Abstraction

Control axis abstraction changes the control inputs to do things other than pure cockpit-centric pitch, yaw, and roll. It is a dynamic mixing dependent on the telemetry of the aircraft relative to the controller's presumed location.

Relative Yaw - Multirotor

Abstraction is perhaps most useful in non-FPV multirotors. A heading is measured from the pilot's position (usually inferred from the takeoff position that the aircraft GPS recorded) to the current GPS position of the aircraft. The pitch axis of the controller is then mapped to 'Away from the controller' vs 'towards the controller', and the roll axis is used to increase tangential velocity. The precise orientations are re-mapped second to second according to the yaw orientation of the multirotor. With an altitude lock enabled, flying a multirotor in this mode can be extremely easy to master, because the mild disorientation that comes with position changing the pitch/roll axes around is removed, and the extreme difficulty of learning to re-map those controls actively is gone. It may be particularly helpful when the multirotor is at a high oblique but not overhead (30-60 degrees up), or when the multirotor is too far away to easily understand orientation. Even for experts who have learned these skills, things like camera work may be a lot easier with a 'free' yaw axis that does not require control compensation.

Relative Yaw - Fixed Wing

In a fixed wing plane, pitch and roll do not lead to simple changes in velocity like they do in a multirotor. Axis abstraction can be used similarly for very distant planes whose orientation is difficult to judge, but it requires a significant guidance/stabilization capability. Relative yaw may be thought of simply as 'Adjust heading hold heading relative to controller-aircraft vector,' so down-pitch might always be used to achieve return to home functionality.

Return To Home

Return to Home is an emergency autopilot function that sends the plane back to its origin, by use of a GPS signal, when wireless communications are lost. This technically requires access to a full range of autopilot-driving sensing, so any system with RTH is likely running on hardware that would support waypoints if they were designed in. This is a frequent addition to OSD/gyro systems on FPV planes, where there is a temptation to fly beyond the radio range, and control is entirely dependent on contact in a high-frequency interference-prone band.

Planned Waypointing

Waypointing provides a flightplan for the aircraft, which flies towards each point on the basis of GPS, 9DOF IMU, and sometimes thermopiles, barometers, and pitot tubes. A tolerance around a destination point is provided in order to provide a realistic success condition. If this tolerance is too small, circling back may occur rather than flying the rest of the waypoints.

In-Flight Waypointing

In-flight waypointing involves assigning a sequence of destinations to a loitering aircraft. The human makes the decision based on something that has changed since the flight began, such as the human's knowledge of video streaming back from the aircraft.

Wind Correction

Wind is a significant problem for waypoint-driven systems, as it throws off the necessary outputs needed to achieve a position. Wind correction solves for some averaged wind vector and transparently tunes its controls to achieve the waypoints.

Automated Takeoff and Recovery

Even in fully autopiloted systems, the two most dangerous parts of flight, launching and landing, are usually human-controlled. It takes a great degree of sophistication to take off and land a fixed-wing plane smoothly along a controlled flight path in varying wind conditions. Recovery is often made plausible in rough terrain with the addition of a servo-triggered parachute.

Autonomous Maintenance

Autonomous maintenance implies the ability to take off, fly for an entire fuel load, land, re-fuel, and take off again for a second flight without user intervention.

Dynamic Waypointing

A decision tree is implemented to take some kind of sensor input, like object recognition performed on a local video feed, and autonomously make decisions about where to go.

Autonomous Navigation

Autonomous navigation involves the ability for the aircraft to avoid flight obstacles unknown to the user.

Sense and Avoid

Sense and avoid is the capability to detect another aircraft passively or based on an active beacon, infer or communicate its flight path, and navigate to a safe distance away from that path in real time without user intervention. It is what traditional aviation authorities want to see in UAV products to ensure safety. A beacon for transmitting UAV position in order to inform an approaching aircraft is often also specified.

Navigation Paradigms

Point to Point

Waypoint navigation usually involves the autopilot directing the plane's nose (or the VTOL craft's pitch/roll direction) in the GPS + magnetometer's idea of the direction of the waypoint. A tolerance around the waypoint is set as a success condition - usually much higher for fixed-wing UAVs - and on success, the next waypoint is attempted. If the tolerance is set too low, circling may occur, though this is minimized in hovering aircraft. This is a simple, resilient system, but the aircraft may not track the line from one waypoint to the next very closely, particularly if that line is long and if there are crosswinds.

Track

A straight track is established from one waypoint to the next, and veering off from this track is corrected aggressively, even if the direction of the craft's heading has to be adjusted to be persistently away from the waypoint. This is ideal for transect imaging/measurements, but also for things like landing on a narrow runway.

Camera Optimized Navigation

Aerial photography can put some particular demands on planes, particularly gliding planes. A camera gymbal, whether one, two, or three axes, can make these goals much easier to achieve, since a plane's feedback loops are linked together with limited degrees of freedom.

Precision Sight

In this mode, the plane tries to hit a particular waypoint at a precise yaw/pitch/roll for a single shot, in order to work as a "flying camera".

Flat Sight

This mode is useful for nadir aerial photography with a fixed camera - the aircraft tries to preserve yaw, roll, and pitch as close to zero as possible.

Circle Sight

The goal of this mode is to maintain fixed camera sights on a target while circling it. This can be very difficult, particularly in a crosswind, even with active gymbals.