Precision Landing
Last updated
Last updated
PX4 supports precision landing for multicopters on either stationary or moving targets. The target may be provided by an onboard IR sensor and a landing beacon, or by an offboard positioning system.
Precision landing can be as part of a , in a landing, or by entering the .
:::note Precision landing is only possible with a valid global position (due to a limitation in the current implementation of the position controller). :::
A precision landing can be configured to either be "required" or "opportunistic". The choice of mode affects how a precision landing is performed.
In Required Mode the vehicle will search for a target if none is visible when landing is initiated. The vehicle will perform a precision landing if a target is located.
The search procedure consists of climbing to the search altitude (). If the target is still not visible at the search altitude and after a search timeout (), a normal landing is initiated at the current position.
:::note If using an offboard positioning system PX4 assumes that the target is visible when it is receiving MAVLink messages. :::
In Opportunistic Mode the vehicle will use precision landing if (and only if) the target is visible when landing is initiated. If it is not visible the vehicle immediately performs a normal landing at the current position.
A precision landing has three phases:
Descent over target: The vehicle descends, while remaining centered over the target. If the target is lost during this phase (not visible for longer than PLD_BTOUT), a search procedure is initiated (during a required precision landing) or the vehicle does a normal landing (during an opportunistic precision landing).
Precision landing can be used in missions, during the landing phase in Return mode, or by entering the Precision Land mode.
0: Normal landing without using the target.
0: Precision landing disabled (land as normal).
Precision landing can be enabled by switching to the Precision Landing flight mode.
:::note When switching to the mode in this way, the precision landing is always "required"; there is no way to specify the type of landing. :::
:::note At time of writing is no convenient way to directly invoke precision landing (other than commanding return mode):
QGroundControl does not provide it as a UI option.
:::note Many infrared based range sensors do not perform well in the presence of the IR-LOCK beacon. Refer to the IR-LOCK guide for other compatible sensors. :::
Precision landing requires the modules irlock and landing_target_estimator. These are included in PX4 firmware by default, for most flight controllers.
IR-LOCK Sensor (external I2C). Disable: 0 (default): Enable: 1).
Landing target is moving (0) or stationary (1). Default is moving.
Horizontal acceptance radius, within which the vehicle will start descending. Default is 0.2m.
Landing Target Timeout, after which the target is assumed lost. Default is 5 seconds.
Final approach altitude. Default is 0.1 metres.
Maximum number of search attempts in an required landing.
Measurement scaling may be necessary due to lens distortions of the IR-LOCK sensor.
If you observe slow sideways oscillations of the vehicle while doing a precision landing with LTEST_MODE set to stationary, the beacon measurements are likely scaled too high and you should reduce the scale parameter in the relevant direction.
To start the simulation with the world that contains a IR-LOCK beacon and a vehicle with a range sensor and IR-LOCK camera, run:
The landing_target_estimator takes measurements from the irlock driver as well as the estimated terrain height to estimate the beacon's position relative to the vehicle.
The measurements in irlock_report contain the tangent of the angles from the image center to the beacon. In other words, the measurements are the x and y components of the vector pointing towards the beacon, where the z component has length "1". This means that scaling the measurement by the distance from the camera to the beacon results in the vector from the camera to the beacon. This relative position is then rotated into the north-aligned, level body frame using the vehicle's attitude estimate. Both x and y components of the relative position measurement are filtered in separate Kalman Filters, which act as simple low-pass filters that also produce a velocity estimate and allow for outlier rejection.
The landing_target_estimator publishes the estimated relative position and velocity whenever a new irlock_report is fused into the estimate. Nothing is published if the beacon is not seen or beacon measurements are rejected. The landing target estimate is published in the landing_target_pose uORB message.
If the target is specified to be stationary using the parameter LTEST_MODE, the vehicle's position/velocity estimate can be improved with the help of the target measurements. This is done by fusing the target's velocity as a measurement of the negative velocity of the vehicle.
Horizontal approach: The vehicle approaches the target horizontally while keeping its current altitude. Once the position of the target relative to the vehicle is below a threshold (), the next phase is entered. If the target is lost during this phase (not visible for longer than ), a search procedure is initiated (during a required precision landing) or the vehicle does a normal landing (during an opportunistic precision landing).
Final approach: When the vehicle is close to the ground (closer than ), it descends while remaining centered over the target. If the target is lost during this phase, the descent is continued independent of the kind of precision landing.
Search procedures are initiated in the first and second steps, and will run at most times. Landing Phases Flow Diagram
A flow diagram showing the phases can be found in below.
Precision landing can be initiated as part of a using with param2 set appropriately:
1: precision landing.
2: precision landing.
Precision landing can be used in the landing phase.
This is enabled using the parameter , which takes the following values:
1: precision landing.
2: precision landing.
You can verify this using the to enter the following command:
only works in missions.
should work, but you will need to determine the appropriate base and custom modes used by PX4 to represent the precision landing mode. :::
The IR sensor/landing beacon solution requires an and downward facing connected to the flight controller, and an IR beacon as a target (e.g. ). This enables landing with a precision of roughly 10 cm (GPS precision, by contrast, may be as large as several meters).
Install the IR-LOCK sensor by following the . Ensure that the sensor's x axis is aligned with the vehicle's y axis and the sensor's y axis aligned with the vehicle's -x direction (this is the case if the camera is pitched down 90 degrees from facing forward).
Install a (the LidarLite v3 has been found to work well).
The offboard solution requires a positioning system that implements the MAVLink . This can use any positioning mechanism to determine the landing target, for example computer vision and a visual marker.
The system must publish the coordinates of the target in the message. Note that PX4 requires LANDING_TARGET.frame to be and only populates the fields x, y, and z. The origin of the local NED frame [0,0] is the home position (you can map this home position to global coordinates using ).
PX4 does not explicitly require a or other sensors, but will perform better if it can more precisely determine its own position.
They are not included by default on FMUv2-based controllers. On these, and other boards where they are not included, you can add them by setting the following keys to 'y' in the relevant configuration file for your flight controller (e.g. as done here for FMUv5: ):
The IR-Lock sensor is disabled by default. Enable it by setting to 1 (true).
determines if the target is assumed to be stationary or moving. If LTEST_MODE is set to moving (e.g. it is installed on a vehicle on which the multicopter is to land), target measurements are only used to generate position setpoints in the precision landing controller. If LTEST_MODE is set to stationary, the target measurements are also used by the vehicle position estimator (EKF2 or LPE).
Other relevant parameters are listed in the parameter reference under and parameters. Some of the most useful ones are listed below.
RTL precision land mode. 0: disabled, 1: , 2: .
and can be used to scale beacon measurements before they are used to estimate the beacon's position and velocity relative to the vehicle. Note that LTEST_SCALE_X and LTEST_SCALE_Y are considered in the sensor frame, not the vehicle frame.
To calibrate these scale parameters, set LTEST_MODE to moving, fly your multicopter above the beacon and perform forward-backward and left-right motions with the vehicle, while landing_target_pose and vehicle_local_position. Then, compare landing_target_pose.vx_rel and landing_target_pose.vy_rel to vehicle_local_position.vx and vehicle_local_position.vy, respectively (both measurements are in NED frame). If the estimated beacon velocities are consistently smaller or larger than the vehicle velocities, adjust the scale parameters to compensate.
Precision landing with the IR-LOCK sensor and beacon can be simulated in .
You can change the location of the beacon either by moving it in the Gazebo Classic GUI or by changing its location in the .
This image shows the as a flow diagram.
