MRDG: Multi-Robots Dataset Generator for CSLAM Backend

Collaborative Simultaneous Localization and Mapping (CSLAM) is critical for multi-robot systems. While frontend development benefits from realistic visual simulators and rich real-world datasets, backend development, lacks of multi-robots, multi-measurements and configurable datasets. This paper introduces a synthetic dataset generator designed specifically for CSLAM backend optimizers. The generator provides fine-grained control over robot trajectories, measurements noises, and, most importantly, inter-robot measurements patterns. We demonstrate the utility of this tool by applying it to compare landmark measurement impact on the accuracy of centralized and distributed CSLAM backends.