The TinyV3RSE Hardware-in-the-Loop Vision-Based Navigation Facility

Abstract

The increase in number of interplanetary probes has emphasized the need for spacecraft autonomy to reduce overall mission costs and to enable riskier operations without ground support. The perception of the external environment is a critical task for autonomous probes, being fundamental for both motion planning and actuation. Perception is often achieved using navigation sensors which provide measurements of the external environment. For space exploration purposes, cameras are among the sensors that provide navigation information with few constraints at the spacecraft system level. Image processing and vision-based navigation algorithms are exploited to extract information about the external environment and the probe’s position within it from images. It is thus crucial to have the capability to generate realistic image datasets to design, validate, and test autonomous algorithms. This goal is achieved with high-fidelity rendering engines and with hardware-in-the-loop simulations. This work focuses on the latter by presenting a facility developed and used at the Deep-space Astrodynamics Research and Technology (DART) Laboratory at Politecnico di Milano. First, the facility design relationships are established to select hardware components. The critical design parameters of the camera, lens system, and screen are identified and analytical relationships are developed among these parameters. Second, the performances achievable with the chosen components are analytically and numerically studied in terms of geometrical accuracy and optical distortions. Third, the calibration procedures compensating for hardware misalignment and errors are defined. Their performances are evaluated in a laboratory experiment to display the calibration quality. Finally, the facility applicability is demonstrated by testing imageprocessing algorithms for space exploration scenarios.

Towards Validation and Verification of Autonomous Vision-Based Navigation for Interplanetary Spacecraft

“…A major concern once a VBN algorithm is designed is how to prove its robustness to hardware error and off-nominal performance. To perform verification and validation, in the space sectors it is possible to use real images from past missions, high-fidelity rendering engines images, or ground-based hardware-in-the-loop (HIL) facility simulations that reproduce space-borne observational conditions [9]. To assess the preliminary performance of the algorithm, a dedicated simulation framework can be implemented to prove the feasibility and to tune the algorithm parameters [10,11].…”

“…On the one hand, the geometric calibration ensures that each object in the field of view is observed as it would be observed in orbit (see Panicucci and Topputo [9] for mathematical demonstration). This is necessary to correct for residual errors in hardware components' alignment and centering and to correct for lens assembly distortions.…”

“…This is necessary to correct for residual errors in hardware components' alignment and centering and to correct for lens assembly distortions. The calibration is obtained as in Panicucci and Topputo [9] by projecting several patterns on the screen and acquiring images with the camera. By knowing the location of points in the screen images and by associating the correct point in the camera image, it is possible to minimize the error between the two LoSes.…”

Hardware-In-the-Loop Validation of Autonomous Interplanetary Navigation Algorithms for Interplanetary Cruises with the Optical Star Stimulator

“…To ensure that the VBN algorithms are both accurate and robust enough to be deployed onboard a spacecraft during an interplanetary mission, it is essential to conduct testing campaigns in a representative environment [17]. These campaigns aim to analyze the performance of the navigation algorithm when real mission conditions and hardware characteristics are introduced in the simulation.…”

“…Therefore, to obtain images with a more realistic error representation, the exploitation of real hardware becomes necessary inside the image generation step. At this aim, optical facilities that include a real optical sensor can be exploited [17,[22][23][24].…”