To Boldly Go Where No Robots Have Gone Before – Part 4: NEO Autonomy for Robustly Exploring Unknown, Extreme Environments with Versatile Robots

摘要

This paper introduces NEO, a novel autonomy framework for controlling a versatile high- degree-of-freedom (DOF) robots such as EELS (a screw-driven snake-like robot), aimed at exploring unknown and extreme environments like the geysers of Enceladus or the subsurface oceans of icy worlds. Distinct from conventional Mars mission strategies, NEO embodies resilience, adaptivity, and risk awareness. NEO supports fault-aware perception using both exteroception and proprioception, inspired by a blind climber’s feat of scaling El Capitan. NEO tightly couples planning, perception, and control, along with leveraging machine-learning- based methods for adaptation. Moreover, NEO incorporates risk-aware decision making with integrated task and motion planning under consideration of uncertainty, enabling autonomous adaptation of actions to mitigate risks and maximize mission success. This paper presents the architecture of NEO, along with experimental results showcasing these capabilities and discusses the potential for NEO in spearheading a new paradigm in space exploration.