Drones have long been buzzing through the air for filming or collecting weather data, for example. Their fields of application will be expanded significantly in the future if the German government has its way. But it is not only unmanned aviation that is in motion. According to experts, the first air taxis with their passengers could populate the skies of major cities and relieve traffic congestion in the next five years. The latest generation of such aircraft, whether air taxis or drones, has long been capable of flying on its own and, in some cases, can also perform complex tasks autonomously. However, it is not possible to consider all safety-related scenarios in advance. Therefore, CISPA faculty Prof. Bernd Finkbeiner is developing the automatically programmable monitoring component RTLola, which can be used to constantly and flexibly monitor the flight capability of autonomous drones during operation.
Without monitoring, unmanned aerial systems can quickly become a safety risk. No matter how smart the systems have become in recent years: Storms, suddenly appearing obstacles, or other flying objects can throw them off their usual course. A crash can have fatal consequences for those moving on the roads beneath the flying objects. "Ensuring that the flying machines can react automatically and safely to all possible imponderables is a major challenge," explains Bernd Finkbeiner. Guaranteeing the greatest possible safety in their operation is an even greater one. The monitoring system ensures that problems during flights are detected early, and appropriate protective measures can be taken.
The system is based on the formal specification language of the same name, RTLola, and consists of several components. "In coordination with the engineers, we first write the specification. That means we exactly determine what properties the system should have and what requirements apply," Finkbeiner explains. Using an algorithm, this formal mathematical description is automatically translated into a logical circuit that can be used to observe flight operations accurately. A hardware component that flies along on the drone picks up raw data from the sensors, such as position, flight altitude, and power consumption, and processes the data further. This enables real-time analysis that provides comprehensive statistics and allows an assessment of the system's health. If the monitor detects from the processed data that requirements such as the minimum distance to an obstacle are being violated, it signals the pilot that there is a problem. In the future, RTLola will also be able to autonomously initiate an emergency landing or return to the drone's base station if a critical situation arises.
"The great thing about RTLola is that our monitor runs completely independent of the control software," Finkbeiner says. That's important because the monitoring component must have access to all the information necessary for the system's overall health but must not interfere with regular operations.
RTLola is already running in test mode in drones at the German Aerospace Center's Autonomous Rotorcraft Testbed for Intelligent Systems (ARTIS), where technologies for autonomous flight are being tested.
translated by: Oliver Schedler