CADRE MoonDB: Distributed Database for Multi-Robot Information-Sharing and Map-Merging for Lunar Exploration

We introduce MoonDB, a distributed database designed to support cooperative robotic exploration and distributed measurements for NASA’s upcoming Cooperative Autonomous Distributed Exploration Rovers (CADRE) mission. MoonDB stores, shares, and fuses information from multiple robots, providing multi-agent planning algorithms with a consistent view of the robotic team. It does so without assuming continuous communication, and significantly limiting bandwidth use through judicious selection of the state variables to share and of the sharing policy and frequency. Further, MoonDB integrates with a pose graph optimization module, allowing mapping information collected by individual robots to be re-localized a posteriori based on refined localization information. Map-merging and reconciliation of inconsistent mapping information uses OpenGL acceleration, resulting in excellent performance on embedded systems. Overall, MoonDB provides a spaceflight-quality solution to the problem of information-sharing for CADRE, addressing one of the key challenges in coordination of multi-agent systems.