Thesis title: Spacecraft and planetary rover navigation through a multi-sensor approach
Space robotic systems have expanded our knowledge about the planet Earth and the other bodies in the solar system, collecting remote sensing and in situ data, carrying out challenging scientific experiments, and enabling access to remote areas on planetary surfaces. To expand the scientific return of the future lunar and planetary robotic missions, novel guidance, navigation and control techniques are currently under development to sustain highly autonomous operations for the next-generation orbiters and surface probes. Autonomous navigation is indeed a desirable requirement in planetary exploration, because interplanetary distances prevent real-time communications between the Earth and the spacecraft.
A main challenge in planetary exploration is the determination of the spacecraft trajectory. An accurate reconstruction of the spacecraft position is a key requirement for space orbiters to retrieve an improved estimation of key geophysical parameters, for planetary landers to support pinpoint landing operations, and for rovers (and helicopters!) to safely drive across the map and reach targets of scientific interest.
To precisely reconstruct the trajectory of deep-space probes, it is fundamental to characterize the dynamical forces acting on the spacecraft. Thermal re-radiation forces, for example, may induce significant perturbations due to the uneven heat emission from the spacecraft bus and appendages. Mismodeling of these effects could have a dramatic impact on the orbit determination process that has been carried out through the analysis of radiometric data acquired from Earth ground stations through single or multi-link established with the deep-space probe. since the beginning of the space exploration era, has been mainly based on radiometric measurements. This data type prevents from supporting real-time operations in critical mission scenarios. Auxiliary data provided by other onboard instruments could be combined with traditional radio-tracking data to improve the spacecraft navigation capabilities, with imaging data representing a promising dataset.
Optical-based navigation techniques have mainly been used in the past decades during the cruise phase of deep-space probes, when the spacecraft is still en route towards the target body, to plan trajectory correction maneuvers while conducting studies of surface topography through limb observations. More recently, onboard imagery has been proven to be very effective to aid in critical close-range maneuvers, such as asteroids terrain sampling or pinpoint landing operations, through the detection and tracking of surface features, such as craters. Also, image-based navigation techniques have been extensively used by planetary roves to aid in the path planning operations, through the detection of hazardous areas, and the estimation of the rover's path through visual odometry techniques. Due to the lack of absolute satellite-based localization infrastructures, visual odometry capabilities are key to limit the accumulation of position errors during traverses on unprepared terrains, enabling the mitigation of slippage-induced errors due to noisy wheel odometry data.
In this dissertation, multi-sensor techniques devoted to the precise navigation of deep-space probes have been designed, implemented, validated and tested by using both synthetic and real data, supporting accurate localization performances across different scenarios, following the ideal route of a probe that orbits around the planet and, once landed, explores the environment navigating towards targets of scientific interest.
The work is organized as follows.
Chapter 1 presents a review of the navigation techniques used to retrieve an accurate positioning of spacecrafts and planetary rovers and describes the motivations of the work.
In Chapter 2 an improved characterization of the dynamical forces acting on planetary orbiters is presented, including thermal recoil effects, which significantly affected the MESSENGER spacecraft at Mercury. The refined dynamical model is used to conduct a reanalysis of the entire MESSENGER's radio-tracking dataset, yielding an improved estimation of the spacecraft trajectory. The results, corroborated by a thorough analysis of the altimetric dataset, indicate that the methodology can be useful to support the orbit determination of the upcoming BepiColombo mission, which is currently on its way to Mercury.
Chapter 3 presents a data-fusion approach, based on the joint processing of radio-tracking and imaging data, to support the orbit determination of the future space exploration missions, through the detection and tracking of surface landmarks via machine vision techniques. The design, validation and testing of the proposed methodology are described in details, including the software pipeline used to generate high-fidelity synthetic images of planetary bodies. The implemented pipeline, based on open-source tools, could be useful to aid in the design of novel vision-based algorithms for pinpoint landing in support of the future surface exploration probes.
In Chapter 4, the implementation of different stereo visual odometry algorithms is presented to support the localization of planetary exploration rovers. Numerical simulations are first carried out to validate the methodologies, and then applied to real images acquired by the Martian rovers, including the rover Perseverance that is currently exploring the Jezero crater, collecting and caching samples of Martian terrain that would be returned to Earth for extensive analyses.
Chapter 5 presents the design, implementation and testing of a stereo vision-based prototype pipeline for the accurate localization of the sample tubes left by Perseverance at the Three Forks depot, in support of the activities of the Mars Sample Recovery Helicopters.
Chapter 6 reports the conclusions of this work.