Hybrid cooperative localization based on robot-sensor networks

Abstract

For cooperative localization in robot-sensor networks (RSNs), efficient utilization of multiple measurements produced by heterogeneous nodes is the key to improving their positioning accuracy. The existing approaches usually focus on the measurement of pseudo-anchor cooperation (PAC) or non-anchor cooperation (NAC), which is often adopted in wireless sensor networks and robotics separately. Therefore, the case where the two types of cooperations are involved cannot be handled reasonably. In view of this, the new concept of a temporary pseudo-anchor (TPA) has been introduced in this work. Owing to the dynamic switching mechanism of TPA, a hybrid range-based node switchable cooperative localization (NSCL) approach can be easily derived, in which the effective PAC and NAC can be realized simultaneously. Furthermore, to extract valuable information from redundant measurements, the TPA selection strategies based on minimum covariance criterion (MCC) and maximum entropy change (MEC) have also been deduced, inspired by the information theory. Benefiting from their different optimization objectives, the two strategies can be employed depending on the requirements of the practical applications. Experimental results indicate that the MCC- and MEC-based NSCL exhibit higher positioning accuracy as compared to the existing PAC and NAC approaches. More importantly, they can still work well even with the only anchor, which exhibits robustness and adaptability to the complex environment.