HISÜL - Highly Secure Intelligent Control and Monitoring of Delivery Drones
short facts
| Duration: | 01.12.2024 - 30.11.2027 |
| Funding: | Bayerisches Verbundforschungsprogramm (BayVFP) des Freistaates Bayern |
| Project partners: | Emqopter GmbH, Gaitd GmbH |
HISÜL develops a highly secure, intelligent control and monitoring system for autonomous delivery drones. Artificial intelligence continuously monitors the drone's condition, detects faults early and intervenes in the control system when necessary. This makes delivery drones more reliable, more efficient and more cost-effective to operate, while reducing operational risk. The solution has two parts: one runs directly on board of the drone and reacts in real time, the other runs on the ground station, where the data from all flights is collected and analysed. The results are to be integrated into existing logistics products and brought to market as a stand-alone product.
Delivery drones can bring medicines, laboratory samples or urgently needed spare parts to their destination within minutes: quietly, with low emissions and independently of road traffic, especially in rural areas. Before they can fly over towns and communities in everyday life, however, they must be as reliable as established means of transport. Today's flight controllers detect only a few predefined faults, such as GPS loss or a low battery. This is where HISÜL comes in: Emqopter, Gaitd and THWS are jointly developing a delivery drone that continuously monitors its own condition, detects faults early, identifies their cause and responds safely, embedded in a fully autonomous delivery chain.
Fully autonomous delivery chain the drone port as a test system
In the drone port, delivery drone, robot and conveyor belt work together without human intervention. The robot picks up parcels from the conveyor belt, loads the landed drone with centimetre precision and unloads incoming shipments. A central unit with server and database coordinates the process. The port thus serves as the interface between the drone and automated logistics and as the test environment for all technologies developed in the project.
A drone that monitors itself
At the heart of the delivery drone is the newly developed control board. It supplies safety-critical components redundantly, switches over automatically in the event of a failure and monitors the health of the batteries. Additional sensors record attitude, rotational speeds, vibrations and control signals. All data flows into a database that serves as the fleet's memory: every flight is an opportunity to learn.
Provoking faults deliberately: real data instead of simulation
Real fault data from drones is rare, because failures are expensive and dangerous. On a test bench, motor failures, controller faults, damaged propellers and external disturbances are therefore induced deliberately and safely. The result is a dataset containing several fault types recorded in a consistent way, something that has hardly existed so far. A second dataset recorded in free flight will follow and is also intended to be made available to the research community.
Safe AI: the best of both worlds.
For fault detection, HISÜL combines artificial intelligence with classical, rule-based methods. Compact neural networks learn exclusively from fault-free flights how the drone normally behaves and report deviations. A rule system based on expert knowledge then derives the cause in a traceable way, for example a damaged propeller or a failed motor. The method thus combines the adaptability of AI with the explainability and reliability of rules.
Mobile delivery chain and emergency landing: the golf cart
Not every delivery ends at a fixed port. An autonomously driving golf cart with a landing platform serves as a mobile test platform for the end-to-end delivery chain of the future. Here we investigate precise landings on moving vehicles, in which drone and vehicle share their perception of the environment from camera, LiDAR, GNSS and IMU (heterogeneous SLAM). At the same time, we are developing methods that enable the drone to identify safe landing sites on its own in an emergency: a key capability when condition monitoring reports a critical fault.
The result is an overall system that makes delivery drones safer, more efficient and more socially acceptable, and that supports safety assessment and regulatory approval with reliable data.
Emqopter's first fully autonomous drone port (photo 1: real installation, photo 2: 3D model). Delivery drones land here, are loaded and unloaded by a robot and take off for the next delivery. The port is the test environment in which the project's technologies are tested and integrated.
Prototype on the test bench: the drone is mounted on a gimbal and can rotate freely but cannot fly away. This makes it possible to provoke and record faults such as motor failures or damaged propellers safely (photo 3). Additional sensors on the arms (photo 4) measure vibrations and rotational speeds. These data are used to develop and test new AI methods for fault detection.
The autonomously driving golf cart (photo 5) complements the stationary drone port as a mobile test platform. A landing platform is being built on its roof (photo 6: concept), on which drones are to land while the cart is moving. Here, methods for dynamic localisation and landing are being developed: a building block for mobile delivery chains.