Positioning with Wireless Signals and Channel Charting

In order to enable reliable data transmission, wireless systems estimate the channel's transfer function (e.g., in the form of an impulse response) using a training sequence that is known to both the transmitter and receiver. The measured CSI contains information about the propagation environment, which can be utilized for wireless positioning. Classical wireless positioning techniques utilize geometrical models for wave propagation in order to perform triangulation or trilateration.  Unfortunately, such approaches require accurate calibration and synchronization among the transmitter and receiver. They also require mostly idealistic line-of-sight propagation conditions to all receive antennas, which often prevents accurate positioning indoors or in dense urban scenarios that contain multiple scatterers. More recently, geometry-free wireless positioning systems that process the measured CSI with the aid of DNNs have emerged. Their principle is simple yet effective: Train a DNN with a dataset consisting of CSI measurements and ground-truth locations obtained via a reference positioning system. Then, use the DNN to process new CSI measurements to generate location estimates. Such positioning systems enable accurate localization even under challenging propagation conditions that prevent approaches relying on geometrical models. The IIP Group has shown that meter-level and centimeter-level accuracy can be achieved for real-world outdoor and indoor scenarios with measured CSI, respectively.

Besides CSI-based positioning, the IIP Group is also concerned with privacy and security aspects that arise from such emerging positioning technologies. Since the extraction of CSI is possible by virtually everyone in the vicinity of the transmitter, CSI-based positioning can be done without the transmitter's (e.g., a mobile user) consent. By extracting location estimates from CSI, privacy is at stake. For example, one can learn the users' behavior (e.g., what stores or bars they frequently visit), their socioeconomic status (e.g., based on where they live and work), or even gain sensitive information about their health (e.g., if the user frequently visits a specialized health clinic). One can even envision concrete security threats, as location information enables one to determine when a user is leaving their home, parking their car, or withdrawing money. The IIP Group is developing novel and effective countermeasures that prevent accurate CSI-based off-device localization without sacrificing quality-of-service.