Motivation
Upcoming space exploration missions envisage precise and safe landing on planetary bodies as well as the execution of rendezvous and docking maneuvers. Such missions will be performed at various distances from Earth, reaching from the vicinity of Earth and Moon out to several astronomical units for targets like Mars, asteroids or the icy moons of Jupiter and Saturn. A reliable execution of the operations mentioned above can only be achieved by implementing spacecraft autonomous operation and utilizing the target body as a navigation reference. This brings a novel requirement for the incorporation of a higher-than-before grade of knowledge about the target, based on a-priori data and on measurements collected during the mission. Examples are globally geo-referenced features originating from mapping missions or the 3D structure of the landing area measured by the spacecraft in the final phase of a landing. This introduces the need for significantly improved online processing capabilities as well as for novel optical sensors.
Optical navigation systems applicable to exploration missions are in the focus of many development projects. Beyond a certain development stage hardware-in-the-loop testing is required for further increasing the maturity. For supporting this kind of testing the DLR Institute of Space Systems built the Testbed for Robotic Optical Navigation (TRON).
Introduction
The Testbed for Robotic Optical Navigation (TRON) is a Hardware-in-the-Loop Test (HiLT) facility, with the purpose to support the development of optical navigation technology. TRON provides an environment which allows qualifying breadboards to TRL 4, and qualifying flight models to TRL 5-6. Typical sensor hardware which can be tested in TRON are active and passive optical sensors like lidars and cameras.
The major components of the lab are a robot on a rail for dynamic positioning of the sensor under testing, a dynamic lighting system for illumination of the targets, laser metrology equipment for high precision ground truth and a dSPACE real-time system for test observation and control, and synchronization of ground truth and sensor data. The laboratory can be customized with user defined hardware such as models of a Lunar or Martian surface. Thanks to this flexibility as well as to its extensive dimensions TRON is well suited for creating scenes representative for the ones encountered by optical sensors during exploration missions.
TRON is already equipped with three scaled models of the lunar surface covering three subsequent sections of a lunar landing trajectory from orbit down to the final approach of the landing site. Furthermore several targets possessing simple shapes such as primitive solids are available. For active sensors such as lidars, the lab is equipped with laser safety curtains allowing the operation of class 4 lasers.
Examples for scenes of exploration missions generated in TRON
TRON Main Building Blocks
1. Dynamics
The dynamic positioning is carried out by a robot system. Its maximum payload amounts to 16 kg on the robot’s hand, and an additional 24 kg to be loaded on the robot’s arm and base. The static repeatability of the robot is about 0.1 mm, its maximum transverse velocity is about 1.5 m/s. The robot is controlled in one of three ways: manually, by programs written in the robot script language, or by the dSPACE real-time system.
2. Terrain Models
Currently three terrain models are installed in the lab.
Terrain model 1 (following the numbering scheme of the images above) measures 9.80 m x 1.96 m. The terrain dynamics, i.e. the range between the lowest and the highest part, is about 6.2 cm. The model has been milled to an accuracy of 1 mm. Its reference data was software generated for being representative of the lunar surface. Due to the self-similarity of the crater size distribution and its definition in the cross-range and down-range altitude reference system it can represent a lunar surface at different scales. As an exemplary use case the robot’s tool center point could be moved along the entire terrain model, varying in distance (above the surface) between approximately 0.5 m and 3 m. Considering terrain model 1 at a scale of 1:10000, the spacecraft position could be simulated over a downrange distance of 100 km and altitudes between 5 km and 30 km. This model served as the terrain model for the breadboard tests of the ESA Lunar Lander project and is currently used for the ATON project. These two projects employ scales between 1:10000 and 1:50000 for the simulation of the Descent Orbit (DO) phase and also the Powered Descent (PD) phase of a lunar landing.
Terrain model 2 measures 3.92 m x 1.96 m and represents a part of the lunar surface in a scale of 1:125000. The terrain dynamics is about 20 cm. The model has been milled to an accuracy of 1 mm, based on reference data derived from the Kaguya mission. Using this model high altitude orbits like the parking orbit as well as the first part of the DO can be simulated. In contrast to terrain model 1 this is truly Cartesian, therefore including the natural curvature of the terrain on the spherical lunar surface. It is used for simulating parts of the DO within the ATON project, as well as for DLR’s research efforts in the field of landmark based absolute optical navigation.
Terrain model 3 measures about 4,20 m x 2,20 m, its terrain dynamics is about 26 cm. The model reference data was obtained entirely by DLR via a process that started with hand-modeling and ended at 3D scanning and post-processing. The model was then manufactured through milling, manually smoothed, painted and equipped with stone-like structures. Due to the hand-made finishing the model is left with a virtually infinite resolution. The model was conceived for the simulation of the last phase of the lunar landing. In this way the terrain relative navigation with respect to the landing site and the evaluation of safe areas can be tested hardware-in-the-loop. Applying low scale factors to this model make it a useful sensor target for 3D imaging sensors. For the ATON project this model is considered to have a scale factor of 1:100.
This landing site model is not only representative of the lunar surface but also of many asteroid surfaces.
Other Sensor Targets
The 3D model of the asteroid 433 Eros possesses an oval-like shape with the dimensions of about 1.0 m x 0.3 m x 0.3 m. The model is a 3D print in scale 1:34000. It is mounted on an axis which allows rotating the asteroid to specific angles and with specific angular velocities.
Another model which is available for customers is comprised by two flat panels sizing together at 2.0 m x 3.0 m. Several primitive bodies can be optionally installed on the panels. This target was used for the characterization of a flash lidar in the FOSTERNAV project.
Optical Environment
The optical environment is simulated via the utilization of a black out system, an anti-reflection system and a lighting system.
The black-out system comprises moving curtains, which cover all windows of the simulations section for isolating the simulations section from external light. Furthermore most surfaces within the simulation section are black for minimizing secondary lighting originating from reflections of internal light sources.
The lighting system comprises a three-axis gantry with the two-axis light source installed on it. The light source is a zoom profile spotlight ADB WARP, using the HMI technique. In combination with the gantry the lamp assembles a 5-DoF system, providing variable solar irradiation angles within the whole simulation section.