Remote magnetic tracking is crucial for navigation in GPS-denied environments, but real-world signals are often too weak, noisy, and complex to model reliably. The authors combine quantum diamond magnetometry with a deep-learning algorithm trained solely on experimental data to track an elevator in an urban setting. They show that full-vector magnetic inputs over short time windows yield robust centimeter-scale accuracy across arbitrary sensor frames despite strong noise, with model training requiring less than an hour. This model-free approach provides a practical route to deploying quantum magnetometry for real-time tracking in complex environments.

Topological phononic systems could enable backscattering-resistant control of mechanical waves for filters and interconnects in wireless and quantum technologies, but they are limited to low frequencies. The authors realize a gigahertz quantum spin Hall effect in nanoelectromechanical phononic crystals. They resolve the opposite handedness of the counterpropagating pseudospin branches in real space, and demonstrate a helical edge channel traversing two 120° corners without measurable loss. This brings topological phononics into the gigahertz regime, opening opportunities for robust phonon manipulation in integrated gigahertz circuits on a CMOS-compatible piezoelectric platform.

Remote magnetic tracking is crucial for navigation in GPS-denied environments, but real-world signals are often too weak, noisy, and complex to model reliably. The authors combine quantum diamond magnetometry with a deep-learning algorithm trained solely on experimental data to track an elevator in an urban setting. They show that full-vector magnetic inputs over short time windows yield robust centimeter-scale accuracy across arbitrary sensor frames despite strong noise, with model training requiring less than an hour. This model-free approach provides a practical route to deploying quantum magnetometry for real-time tracking in complex environments.