ESA GNC Conference Papers Repository

Simulating planetary approach and landing to test and verify autonomous navigation and guidance systems
I.M. Martin, M.N. Dunstan, S.M. Parkes, M. Sanchez-Gestido, G. Ortega
Presented at:
Salzburg 2017
Full paper:

Vision systems can be used in space applications to perform absolute navigation, surface relative navigation and hazard detection to support safe and precise landings but validation of these mission critical vision-based navigation systems is difficult [1]. A simulated test environment that generates realistic approach and descent images can play an important role in testing, developing and validating these systems. Providing realistic simulated images at high speed with model resolution ranging from kilometres to centimetres is a challenging task due the large model sizes, the image resolution range and the requirement to support closed-loop testing in real or near-real time. PANGU (Planet and Asteroid Natural scene Generation Utility) is a software suite specifically designed to test, verify and validate a wide range of vision-based space applications such as planetary landers, small-body approach and landing and surface rover applications by generating simulated vision and LiDAR images of planetary surfaces and asteroids. The images are generated from multi-resolution models which are created from a combination of real and artificial data. A complete simulation from approach to landing can be run from a single model rendered at near-real time (e.g. 10 Hz) to generate images for closed-loop testing. The custom renderer implements Bidirectional Reflection Distribution Functions (BRDFs) such as Hapke, Oren-Nayar and Lunar Lambert, which can be tuned to match the reflection properties of specific planetary surfaces [2], atmospheric effects and a full, parameterised camera model enabling realistic images to be generated for a wide variety of mission scenarios. PANGU models can be based on real terrain elevation data which is synthetically enhanced to the required resolution in a flexible multi-resolution model. Whole planetary bodies or sections of planetary surfaces can be defined from Digital Elevation Models (DEMs) which are 2D maps of height values in various standard projections. The resolution of the model along the descent trajectory region can be artificially enhanced by specifying regions of increasingly higher resolution towards the target landing site. Multiple higher-resolution landing sites can be defined if required. The resolution is enhanced by combining terrain interpolation with additional fractal detail and adding small-scale features such as craters, boulders, positive relief features or dunes in user-specified size-density distributions. The advanced crater model incorporates customisable crater degradation and rim irregularity distributions, models overlapping craters and can realistically integrate craters into the existing terrain [3]. Asteroids and other small bodies are typically highly irregular in shape and usually cannot be represented as DEMs. The Implicitly Connected Quadrilateral (ICQ) format is being widely used to represent publically available shape models of asteroids and other small bodies [4]. These models are usually low-resolution but PANGU can be used to import these shape models, artificially increase the resolution of the whole model by combining the shape model with customisable fractal terrain noise and add distributions of craters and boulders below the resolution of features defined in the original shape model. This makes it possible to create a high-resolution model of a small body starting from a low-resolution shape model, using PANGU to enhance the resolution and create a multi-resolution landing site region. This model can then be used to test and validate the navigation approach and landing systems of future missions to specific small bodies. PANGU simulations can be run with multiple planetary bodies, imported CAD models of spacecraft, satellites or surface rovers and appropriate lighting and reflection models. Pre-defined trajectories and model dynamics can be defined for open-loop tests. PANGU can also be run as a server to provide images for closed-loop simulations in response to TCP/IP socket commands that can be integrated with other systems such as Matlab, SIMULINK or custom built applications calling PANGU C or Java APIs. Simulations can also be integrated with the NAIF/SPICE system [5] to enable publically available ephemerides to be used for accurate simulations of real world historical as well as predicted future events allowing SPICE to control the position and orientation of the Sun, the camera, planets and asteroids in the simulation. This enables an accurate rendering of a given view to be generated based only on the time of the event. It has been used, for example, to recreate AMICA images of asteroid Itokawa taken by Hayabusa. PANGU contains a sophisticated, fully parameterisable, camera model which models the transfer of photons through the optical system to the detector, the conversion to photo-electrons and then ADC (Analog-to-Digital) conversion to a digital number. It has realistic models of photon and electron shot noise, read-out noise, radiation, saturation and blooming, and various per-pixel non-uniformities. The camera model uses the GPU on the graphics card for maximum performance. It can operate on larger radiance images than existing software implementations typically use and usually needs less than 50ms to compute the results. The effects of radiation, due to protons and electrons, is modelled in detail to support simulations in different radiation environments such as at Jupiter or around the Earth. Other noise sources are also modelled including thermal dark current as a function of temperature and exposure time, and read-out noise. The camera model has been validated against the Clementine UVIS and Haybusa AMICA cameras and has been used to model other cameras such as MSL/MARDI and MEX/VMA. PANGU has been verified and validated at unit and integration level as well as through a series of absolute validation tests comparing the PANGU outputs against images taken with real cameras using real world elevation and shape models. PANGU is able to verify and validate vision-based navigation systems throughout the design and qualification phases of ESA space programmes. It can be used to support landing missions on the Moon, Mars and asteroids using realistic surface and camera models. The full paper will describe the new, recently developed PANGU features including the creation of the whole planet and asteroid models with example mission scenarios, the camera model and the verification process. PANGU was developed by the University of Dundee for ESA and is being used on many European activities aimed at producing precise, robust planetary lander and rover guidance systems.