2nd International Mars Sample Return 2018 (LPI Contrib. No. 2071) 6126.pdf

CNES ROVER AUTONOMOUS NAVIGATION AND ITS POTENTIAL APPLICATION TO MARS SAMPLE RETURN FETCH ROVER. M. Delpech, X. Rave, L. Rastel. 1Centre National d’Etudes Spatiales (CNES), 18 Avenue Edouard Belin, 31400 Toulouse, France - email: [email protected]

Introduction: Rovers on the Mars surface have re- the consecutive local maps taken at each stop and such lied over the years on navigation approaches with pro- a representation leads to range limitations due to obvi- gressively higher autonomy However, the distance ous on-board memory constraints. crossed by Curiosity today does not exceed several tens To increase the rover long traverse capability, of meters per sol and the daily roving distance of ESA CNES developed two additional features (Regional Exomars, expected to land in 2021, should be in the map build & update, Regional path planning) that rely 100 m ballpark. The future MSR mission is now push- on a more efficient and complementary representation ing this boundary with the concept of fetch rover which of the environment. To optimize memory usage, the traverse capability should be increased by an order of local traversability map generated at each perception magnitude to optimize samples retrieval [1]. This paper cycle is compressed into an obstacle map that can be presents the work achieved by CNES to meet some of vectorized. Such a compact representation referred as these challenges throughout its R&D activities and regional map stores all identified obstacles along the participation to Exomars. way and provides a synthetic and global knowledge compatible with few kilometers traverses. CNES activities in AN : The history of CNES ro- The regional map enables the rover to efficiently plan a botics team involvement in rover autonomous naviga- reverse path if dead ends are encountered or a quick tion dates back to the mid 90s with its participation to return is required. In addition, the regional map can be the Russia led Marsokhod project. CNES contribution initialized with forbidden or undesirable areas labelled comprised the stereo camera system and the Autono- as such by ground operators. These functionalities that mous Navigation (AN) software package which opti- have a low impact in terms of memory usage and com- mization required intense effort due to the limited putational cost with respect to the typical 3 stages AN computing resource. AN algorithms were subsequently algorithms have been developed to comply with Exo- improved and matured through several years of R&D mars requirements. The computational time of the development and validation effort including field test- whole optimized AN package on the Exomars LEON2 ing. This expertise was proposed to JPL during the first processor is around 30 s for a sequence of 3 percep- MSR project and experiments/demonstrations were tions. These functionalities that can be activated or performed in JPL Mars Yard in 2000 to show the po- inhibited from ground are foreseen to be exercised du- tential of Autonomous Navigation versus the nominal ring the nominal and/or extended mission to show their Hazard Avoidance approach. Afterwards, CNES got potential in the perspective of longer traverses. This involved in the ESA-led Exomars project through an AN technique has been intensively validated on a re- active participation in the requirements phase, the in- presentative numeric simulator and tested on the field kind delivery of its AN framework to the main actors with ARTEMIS rover (Exomars-like rover platform). along with its technical support[2]. Following the stall Further improvement is currently under study to exploit of the AN functionality due to budget constraints, the regional map concept for relative localization. CNES recently stepped forward to promote the re- introduction of AN in Exomars and is currently negoti- Conclusion: This work describes a potential step to ating the accommodation of its own AN flight software enhance the rover autonomous navigation that repre- along with the UK AN baseline solution. The CNES sents only one of the challenges to be faced by the proposed functionalities go beyond the strict Exomars MSR fetch rover. Besides the necessary ruggedization requirements and include capabilities that appear bene- of the locomotion platform, the implementation of reli- ficial to future rovers with enhanced autonomy needs. able perception approaches for the detection, the local- ization and the grasping of sample tubes will require Regional Navigation: The typical structure of a significant effort. Vision Based AN package relies on 3 sequential com- putational stages performed while the rover is stopped: References: [1] Planetary Science Decadal Survey, (1) Stereo Correlation & DEM Generation, (2) Traver- MSR Lander Mission, P. Christensen, L. May, NASA sability Map Build & Update, (3) Path & Perception April 2010. [2] Autonomous navigation: a develop- Planning. There, the environment representation used ment roadmap for Exomars. L. Rastel, M. Maurette, for short-medium range planning is built by merging ASTRA 2006.