Global Single-Shot Indoor Localization of Autonomous Flight Robots With Minimal Sensor Configuration

This paper introduces a novel localization system for global single-shot localization of flight robots in indoor environments. The system is designed for initialization at standstill, to ensure that a safe and hazard-free state is confirmed before the engines start. Existing localization solutions are often unsuitable for intralogistic applications because of to their weight, energy consumption, or infrastructure costs. The proposed system achieves global localization with minimal sensing hardware. The pose can be reliably determined using a few 1D distance sensors and a digital map of the environment. Two methods are introduced, adapted to perform single-shot localization with resource-constrained sensor configurations. Both methods use application-specific knowledge to initially reduce the search space during localization. Afterwards, one of two methods is employed: a randomized coarse-to-fine search, in which the search space is iteratively reduced, or a parallel differential evolution algorithm, which enhances pose estimates with evolutionary algorithms while avoiding premature convergence. In a real scenario, the system with six 1D sensors and an inertial measurement unit achieved a 100% success rate in localization, with position errors below 0.1 m and rotation errors below 0.05 rad. Owing to the low information density of sensor data, transmission to external edge computing hardware is efficient and enables remote processing with minimal bandwidth requirements. This reduces onboard computations, system weight, and energy consumption of the flight robot. The localization system provides a lightweight and energy-efficient solution for global pose estimation in flexible productions and enables the economical use of autonomous flight robots in industrial intralogistics.