Accurate and Reliable Mobile Robots Localization Based on Secondary Radar and Multi-Modal Sensor Fusion

Abstract

Service robots are expected to rapidly grow in spread and importance in the years to come. Service robots are autonomous mobile robots performing useful tasks for humans, excluding manufacturing operations. Thus, they frequently operate in unstructured and dynamic indoor environments among humans. A crucial prerequisite for the efficient and safe operation of such devices is the reliable and accurate localization. Although widely spread for global positioning and navigation, satellite-based technologies such as GPS cannot provide sufficient reliability and accuracy in many outdoor and virtually all indoor scenarios. As service robots frequently operate indoors, an alternative is needed. Many promising approaches have been demonstrated in the past, however, no technology has yet managed to reach mass-market adoption for indoor localization. The aim of this work is to present a powerful alternative for indoor and outdoor local positioning based on secondary radar combining range and angle measurements. This combination allows for accurate and reliable localization with much less infrastructure compared to state-of-the-art multilateration and multiangulation techniques. This is thoroughly studied based on a frequency-modulated continuous-wave (FMCW) single-input multiple-output (SIMO) secondary radar. The theoretical limits on the achievable measurement accuracy are shown and analyzed in multiple experiments. The effect of intrinsic and extrinsic error sources on range and angle estimation accuracy is also studied. An indoor localization system for healthcare service robots is developed using the proposed radar device. Using multi-modal sensor data fusion with an ultrasonic device and odometry, accurate and robust positioning is achieved and verified in three real-life scenarios. Another system for 6-degrees-of-freedom (DOF) service robot localization for planetary exploration missions based on the same concept is analyzed theoretically in terms of achievable accuracy. Measurement campaigns in three different scenarios show the advantages of the approach to combine range and angle measurements for 6-DOF localization. The versatility and flexibility of the proposed concept is also proved by a successful participation in the Microsoft Indoor Localization Competition 2016 and a UAV 3D localization experiment demonstration.