Indoor robots, in particular AI-enhanced robots, are enabling a wide range of beneficial applications. However, great cyber or physical damages could be resulted if the robots’ vulnerabilities are exploited for malicious purposes. Therefore, a continuous active tracking of multiple robots’ positions is necessary. From the perspective of wireless communication, indoor robots are treated as radio sources. Existing radio tracking methods are sensitive to indoor multipath effects and error-prone with great cost. In this backdrop, this paper presents an indoor radio sources tracking algorithm. Firstly, an RSSI (received signal strength indicator) map is constructed based on the interpolation theory. Secondly, a YOLO v3 (You Only Look Once Version 3) detector is applied on the map to identify and locate multiple radio sources. Combining a source’s locations at different times, we can reconstruct its moving path and track its movement. Experimental results have shown that in the typical parameter settings, our algorithm’s average positioning error is lower than 0.39 m, and the average identification precision is larger than 93.18% in case of 6 radio sources.

中文翻译：

---

面向安全的室内机器人跟踪：对象识别观点

室内机器人，尤其是人工智能增强型机器人，正在实现广泛的有益应用。然而，如果机器人的漏洞被用于恶意目的，可能会导致巨大的网络或物理损害。因此，需要对多个机器人的位置进行连续的主动跟踪。从无线通信的角度来看，室内机器人被视为无线电源。现有的无线电跟踪方法对室内多径效应很敏感，而且容易出错，而且成本很高。在此背景下，本文提出了一种室内无线电源跟踪算法。首先，基于插值理论构建RSSI（接收信号强度指标）图。其次，在地图上应用了 YOLO v3（您只看一次版本 3）检测器来识别和定位多个无线电源。结合不同时间源的位置，我们可以重建其运动路径并跟踪其运动。实验结果表明，在典型参数设置下，我们算法的平均定位误差小于0.39 m，在6个射源的情况下平均识别精度大于93.18%。