Extending ROS Transform Library for Massive Autonomous Robots

TF is a core library in ROS that manages positional relationships between objects. TF has fundamental problems in that accesses are sequentially executed due to the giant locking and its interfaces' limited functionality. TF is designed for managing only a few objects, and it has two problems supporting such use cases as pedestrian prediction or traffic congestion avoidance. The first problem is poor operators, and the second problem is the generation of phantom anomalies. We address the first problem by designing a SQL-based interface with various powerful operators and the second problem by designing a phantom anomaly protocol based on ghost record protocols.