ETH Zürich – Institute for Dynamic Systems and Control
Flying Machine Arena
The Flying Machine Arena (FMA) is a portable space devoted to autonomous flight. Measuring up to 10 x 10 x 10 meters, it consists of a high-precision motion capture system, a wireless communication network, and custom software executing sophisticated algorithms for estimation and control.
The motion capture system can locate multiple objects in the space at rates exceeding 200 frames per second. While this may seem extremely fast, the objects in the space can move at speeds in excess of 10 m/s, resulting in displacements of over 5 cm between successive snapshots. This information is fused with other data and models of the system dynamics to predict the state of the objects into the future.
Empfohlene Produkte
Vicon Vantage
Die Vantage-Serie stellt das Flagship von Vicon dar. Die Auswahl erstreckt sich von 5 MP- über 8 MP- bis zu 16 MP-Kameras bei Aufnahmeraten von bis zu 2000 Hz. Verschieden Onboard-Sensoren erleichtern die tägliche Arbeit mit dem Systems. Für Aussenaufnahmen, grosse Aufnahmevolumen und besonders schnelle Bewegungen sind diese Kameras die erste Wahl.
Vicon Vero
Die kompakten Vero-Kameras gibt es als 2.2 MP und 1.3 MP Variante. Dank Vario-Objektiv und Sensor im 2:1 Format (Vero 2.2), lassen sich diese Kameras sehr vielseitig einsetzen und jederzeit flexibel an verschiedene Aufnahmevolumen/Räume anpassen. Einzigartig ist auch das Preis-Leistungs-Verhältnis der Vero-Serie.
Tracker
Vicon Tracker ist mit dem Hauptziel entwickelt worden, so schnell wie möglich rigide Objekte zu tracken und deren Positionen und Ausrichtungen an andere Software weiterzuleiten. Deswegen ist die Latenzzeit sehr kurz – bis zu 2.5 ms. Tracker überzeugt auch mit ihrem Algorithmus, der ein Object sogar nur mit einer Kamera weiter tracken kann, was das Trackingvolumen für einen bestimmten Kamera-Aufbau maximiert.
Our Flying Machine Arena System
A mobile version of the Flying Machine Arena system was built for performances outside the lab. Physically, this is a set of laptops and a standard 19 inch rack, housing two Vicon Giganet hubs for up to 20 Vicon cameras and a managed smart switch. The cameras are packaged in dedicated cases and the rack is shock-proofed, resulting in approximately twelve ship-ready cases, with a total weight of approximately 300 kg. When running with 19 cameras, the total system consumes approximately 600 W.
The current system has been used for flight spaces ranging in size from 185 m3 to 720 m3 . The size of these spaces is largely defined by the available space at the venue in conjunction with the possible camera mounting points. In order to function properly, the motion capture cameras must remain rigidly fixed relative to each other, and, preferably, relative to the space. A good solution for this is a metal truss suspended from the ceiling, though other solutions, such as mounting the cameras directly to mount points on the ceiling/walls, have been used. Prior to arriving at the venue, a camera simulation tool is used to predict the space coverage of the motion capture system. The user then iterates over the camera positions and orientations until a satisfactory result is obtained, and defines the flight space based on the space coverage results. The truss may also serve a dual purpose as a mounting point for safety nets, though in this case two trusses should be used to avoid people on the ground from moving the camera-truss assembly.
Given a prepared suspended truss structure for the cameras and nets, it currently takes four people approximately 6 h to properly set up the mobile system, though this estimate depends on the specifics of the setup. After the initial set up of the system, a calibration of the motion capture system is performed using a calibration wand (provided by the manufacturer of the motion capture system), mounted on a long, telescopic rod to allow the user to move the wand through the entire flight space. The calibration algorithm computes, amongst others, the position and attitude of each individual camera. This data is then fed back into the camera simulation tool to verify the predicted space coverage, and adjustments are made to the camera placement if necessary.
