Autonomous Science Target Identification and Acquisition (ASTIA) for planetary exploration

Abstract

We introduce an autonomous planetary exploration software architecture being developed for the purpose of autonomous science target identification and surface sample acquisition. Our motivation is to maximise planetary science data return whilst minimising the need for ground-based human intervention during long duration planetary robotic exploration missions. Our Autonomous Science Target Identification and Acquisition (ASTIA) architecture incorporates a number of key software components which support 2D and 3D image processing; autonomous science target identification based upon science instrument captured data; a robot manipulator control software agent, and an architecture software executive. ASTIA is being developed and tested within our Trans-National Planetary Analogue Terrain Laboratory (PATLab). This provides an analogue Martian terrain, and a rover chassis with onboard manipulator, cameras and computing hardware. Experimentation results with ASTIA and our PATLab rover are presented.

Effective Feature Selection for Mars McMurdo Terrain Image Classification

“…This, in combination with the observation that both IGR and FRFS preserve the underlying semantics of the selected features, also indicates that information loss can be minimized and even avoided in building the classifiers. Such work is of particular significance for on-board classification and analysis of large-scale images in future Mars rover missions [1].…”

“…principal component analysis [3]), without destroying the underlying semantics of the features. This is of great importance to on-board image classification in future Mars rover missions [1].…”

Multi-modal active perception for information gathering in science missions

“…More recently, KSTIS, a fuzzy, knowledge based expert system was developed to emulate how a planetary geologist would assess a scene [36]. Although fuzzy systems provide more sophisticated identification of scientifically valuable targets than weighted functions, they require careful synthesis of rules to be effective, which is nontrivial in remote and previously unknown environments.…”

A probabilistic approach to autonomous planetary science with prime and scout rovers

“…The work described above has recently evolved and is now known as the Autonomous Science Target Identification and Acquisition (ASTIA) architecture [30]. In this architecture, science target identification is improved through a new knowledge-based approach, which uses rule-based fuzzy logic to classify structure, texture and composition.…”