There is a newer version of the record available.

Published October 18, 2023 | Version v1
Conference paper Open

PerSim: Perception for Planetary Prospection and Internal Simulation

Creators

  • 1. ROR icon German Research Centre for Artificial Intelligence

Description

For planetary robotics autonomous prospecting, robust, long-term navigation becomes crucial. The goal of the research project PerSim is to develop technology to address some of the challenges of 
active perception for resource identification and long-term navigation strategies in an integrated architecture.
The fist assessment addressed autonomous selection of regions for inspection, combined arm-base approach, close range data acquisition and categorization of the acquired spectral data using Deep Learning.
Furthermore, autonomous navigation including potential failure prediction and avoidance are also scoped.
The following targets are pursued in the second assessment: an internal simulation to enhance the system safety and provide means for autonomous onboard safe testing, an episodic memory representation to serve as basis for the implementation of long term adaptation and finally a repertoire of behaviors to enable different motion modalities.
The paper provides insights on the approaches and initial results.

Files

14359_persim_astra_2023.pdf

Files (55.0 MB)

Name Size Download all
md5:89a007253c02971545cde88afc257370
55.0 MB Preview Download