Introduction
Narrowband ultrasound (NBU) is a cornerstone of high-precision indoor positioning due to its low cost and centimeter-level accuracy. However, NBU systems face a fundamental "speed limit" imposed by the speed of sound (~340 m/s). Traditional protocols like ActiveBat or Cricket rely on Time Division Multiple Access (TDMA), where each target is assigned an exclusive time slot to prevent signal collisions. As the number of targets increases, the update frequency for each individual device drops significantly, making it difficult to track fast-moving objects in multi-agent environments.
To overcome this, UltraChorus introduces a "chorus" ranging paradigm. Instead of waiting for exclusive slots, multiple targets transmit simultaneously. The system then disentangles these overlapping signals using geometric constraints and probabilistic filtering.
The UltraChorus Architecture
Physical Layer and the Blind Region
The primary challenge in simultaneous transmission is the "aftershock" effect—the mechanical inertia of a receiver's transducer that keeps it active after a pulse ends. If two signals arrive too close together in time, they merge into a single undecipherable waveform.
UltraChorus defines a "confident separation distance" ($\omega$). If the distance difference between two targets and a receiver is greater than $\omega$, their signals can be distinguished. This physical constraint creates a "Blind Region"—a spatial area where one target's signal effectively masks another's at a specific receiver.
Resolving Anonymity and Ghost Positions
Since NBU pulses do not carry unique identifiers, receivers obtain a "bag of distances" without knowing which distance corresponds to which target. UltraChorus solves this data association problem through two consistency checks:
- Historical Consistency: New distance measurements are compared against a target's last known position and motion model.
- Self-Consistency: The system verifies if a set of distances from different receivers intersects at a single point (minimizing residue) to filter out "ghost" positions.
For complex scenarios where targets cross paths, a Probabilistic Particle Filter tracks multiple potential trajectories. It evaluates the likelihood of a target's velocity and acceleration to "un-mix" signals that might be temporarily obscured.
Location-Based Time-slot Scheduling (LBTA)
When targets are physically too close to be separated by geometric logic, the system dynamically assigns them to distinct time slots. This hybrid approach ensures that UltraChorus maintains reliability even in high-density clusters, gracefully reverting to TDMA-like behavior only when necessary.
Evaluation and Results
The system was evaluated using both a custom MATLAB simulator and a hardware testbed featuring modified MTS450CA Cricket nodes.
- Accuracy: In simulations with 10 targets, 90% of locating errors were under 1cm. In real-world hardware tests involving a toy train moving at 1m/s, the system maintained a 90% error rate of less than 15cm.
- Update Rate: UltraChorus significantly increased locating capacity. While traditional exclusive modes would severely limit 10 targets, UltraChorus allowed an average of 8 targets to be located within a single time slot.
Conclusion
UltraChorus demonstrates that the "speed of sound" bottleneck in ultrasound positioning can be mitigated by shifting from exclusive time slots to simultaneous "chorus" transmissions. By leveraging the physical properties of transducers and robust geometric filtering, the system achieves high-frequency updates for multiple targets without sacrificing the centimeter-level precision that makes ultrasound attractive for indoor localization.
Summary updated