How to reduce errors during the launch of drone swarm to ensure that the fleet always takes off with full force

Written by
Reading time:
min
Drone Show Technologies
August 4, 2022

Alex Levandovskiy the head of the Drone Show Software division at SPH Engineering about alerts function.

Alerts is a functionality in the software that allows you to reduce the likelihood of emergency situations in the air due to a defect of the drone. This function works in two stages:

  • at the preparation stage, it allows to determine that the drones are not ready to fly and the reason for this,
  • at the fleet flight stage, it allows to identify a “bad drone” and remove it from the drone dance.

A drone for the show has two levels of flight readiness, physical and logical. Physical is the level of the autopilot itself - readiness and correct calibration of sensors, EKF convergence, GPS accuracy. The logical level is the presence of a loaded dance, correctly set speed, center and start time of the dance, as well as time offsets for takeoff, correctly filled geofences (a technology that allows you to set the boundaries of the territory) and other parameters.

Until recently, there were often situations when the operator, having put the drones on the field, found out that some of the drones were not ready to fly, either physically or logically.

In order to understand the reason for the logical unavailability, values were transmitted in telemetry from the drone that described the current logical state of the drone. But the condition of the drone was assessed only in general terms and without any specific numerical values. For example, the drone signaled that it would not have time to fly to the starting point, but it was impossible for the operator to understand details of time shortage and what parameters should be changed.

With the cause of physical unpreparedness, the situation was even worse. The operator was simply shown the fact that the drone was not ready to fly, and the operator could find out the reason from the server text logs, which might not reach the base station in case of connection problems. Working with text logs is quite inconvenient in the case of large fleets, since hundreds or even thousands of information blocks must be processed.

First of all, the functionality of alerts was integrated to solve exactly those problems and difficulties that are listed above, since they were identified “in the fields” and in practice.

Functionality of alerts

The alert functionality is an advanced mechanism for notifying the operator of problems with the confirmation from the ground in order to ensure the delivery of a message in case of poor connection quality. The message contains a coded description of the problems the drone has (or lack thereof) with the necessary explanatory numerical values.

For example, returning to the situation described above, the operator may be told that the drone will not have 5 seconds to reach the start point of the dance, and the operator will be proposed to change the time offset parameter, or the parameter of the movement speed to the start of the dance.

The same applies to problems with sensors (accelerometers, etc.), the operator is informed which specific sensors are currently producing low-quality information so that the operator can recalibrate the sensors or decide to replace the drone.

If problems with the logical readiness level are solved strictly at the stage of preparation for the show, then alerts to the physical readiness level can also occur during the flight: fast battery drain, problems with GPS, increased vibration due to a strong gust of wind and, as a result, loss of positioning accuracy and other. Such alerts have several levels of severity: warning, error, critical, emergency.

The operator does not see warning level alerts during a flight, but can analyze their presence after the end of the flight, but can analyze their presence after the end of the show. All other levels of alerts are displayed in front of the operator in a special window in a form that will allow the operator to quickly understand the essence of the problem, including getting an answer to the question of whether a critical situation has arisen for one drone or for a significant part of the fleet. Then the operator decides to stop the show or to withdraw specific drones from the show.

Our task is to give the operator a convenient and reliable tool for this. Of course, the quality of the connection affects the operation of this tool.

At the moment, we have implemented logic for 90 different alerts. This functionality delights everyone who already uses it by updating the version of Drone Show Software to 4.1.

About