The Promise and Challenge of Deploying Safe Artificial Intelligence on Manned Planetary Exploration Rover

In recent years, a multitude of countries and space agencies, including China, the United States, Russia, Japan, and the European Space Agency (ESA), have made significant announcements outlining their ambitious plans for manned moon landings and the establishment of lunar bases. For long-duration and long-distance explorations, the manned planetary exploration rover (MPER) is an essential equipment to make the missions feasible and greatly improve the efficiency. However, the harsh environments such as low gravity and rigid-flexible coupling terrain on the surface of the moon/Mars make it a huge challenge to ensure the safety and stability of high-speed motion. Thus, the development of safe artificial intelligence that supports remote operation, on-orbit driving, and automatic driving has become an ideal solution. Deploying safe artificial intelligence on MPER and achieving the three working modes require a high degree of collaboration of the latest technologies in the fields of perception, planning, and control. In detail, the MPER can provide real-time multi-source perception information to assist the operator in decision-making, and realize the safety and stability of the underlying control when working in the remote operation or the on-orbit driving mode. As for the automatic driving mode, the MPER can independently plan and make decisions based on macro-task objectives and execute them. By applying the latest research results to MPER, a profound technological revolution awaits the field of planetary exploration. However, there are still some challenges that need to be addressed regarding the specific application of these technologies. For example, when it comes to perception, factors such as large sensor chattering and short-term differences in sensing information under high-speed movement lead to blurred and distorted sensing images or low signal-to-noise ratio of sensing data. These problems bring difficulties to wheel condition monitoring and map construction that base on sensor information. Besides, trajectory drifting during turning and vehicle jolts over obstacles enhanced by low gravity are problems that cannot be ignored for control. This paper proposes possible solutions to these challenges, providing a reference engineering scheme for the upcoming manned lunar exploration of China.