Accurate indoor drone four degrees-of-freedom (4DoF) estimation, including both 3D position and heading, is essential for safe flight in GNSS-denied environments, opening up opportunities for fully autonomous IoT applications such as infrastructure inspection and surveillance. Approaches using a built-in camera are attractive due to their low cost and ease of deployment, however, existing methods face key limitations. Visual-Inertial Odometry suffers from drift and often fails in textureless environments, while Visible Light Positioning requires a direct line of sight to ceiling LEDs, which is challenging for drones with a limited field of view. We propose ReFDrone, a system that simultaneously estimates 3D position and heading from a single image using a downward-facing rolling-shutter (RS) landing camera and floor-reflected light from modulated ceiling LEDs. Our core insight is that RS distortion produces spectral lobes in the frequency domain whose orientation directly encodes the angle of arrival (AoA) of the LED. Leveraging this property, ReFDrone extracts AoA from a single RS image, enabling estimation of 2D position and heading from just two LEDs, and derives altitude using a virtual-view image with a distance model. We evaluated ReFDrone under 1296 conditions in 2.5 m x 2.5 m testbed, achieving a 90th percentile 3D position error of 0.335 m and a heading error of 8.101 degrees, demonstrating its accuracy and practicality.

Beamforming with speaker arrays allows for directional audio communication, selectively delivering information to users in specific locations. However, this approach typically requires a specialized array hardware. We propose an adaptive acoustic spot communication method targeting moving receivers that utilizes only two commercial off-the-shelf (COTS) speakers. A key feature of our method is its ability to achieve high-performance by predicting the receiver’s position. We conducted real-world experiments with a moving microphone to evaluate the effectiveness of the proposed method. The results showed a demodulation success rate of 94.7% for the adaptive spot communication, excluding the region directly in front of the speakers, demonstrating the feasibility of adaptive spot communication for a moving target.

Smartphone positioning utilizing a single indoor speaker and its reflections offers the advantage of low deployment cost. However, unlike the direct wave, reflected waves are not always stably detectable, even under Line-of-Sight (LOS) conditions, leading to inconsistent positioning performance. Additionally, associating each detected reflection with its corresponding virtual speaker poses a challenge. This study proposes a method to improve positioning performance by tracking the smartphone based on hypotheses about the identity of each detected reflection. Evaluations were conducted in a real-world environment along both rectangular and circular paths, using only the smartphone’s bottom built-in microphone. The results demonstrated 90th percentile errors of 0.62 m and 0.77 m for the rectangular and circular paths, respectively. This accuracy, approximately equivalent to a step length, suggests potential for navigation applications. These results represent error reductions of 55% and 46%, respectively, compared to conventional approaches.

This study investigates an indoor 3D positioning method for a microphone using a single speaker. The proposed method estimates the distance, azimuth, and elevation of the microphone from the speaker and calculates the 3D position of the microphone. Multiple short signals with different frequency bands were transmitted sequentially at sufficient intervals to avoid the influence of indoor reflected waves. We extracted these signals at the receiver side, and their amplitude spectra were computed. The azimuth and elevation of the microphone were estimated by matching the obtained amplitude spectrum with the reference data previously measured at each azimuth and elevation. The distance was estimated from amplitude attenuation. The 3D positioning performance was evaluated at ten points in a practical environment. The results show that the 90-th percentile value of the error was 0.583 m.

This paper proposes a tracking system that integrates visible-light positioning and pedestrian dead reckoning using a smartphone’s built-in ambient light sensor and inertial sensors. By correcting the accumulated errors in the position estimation via ambient light sensing, the proposed system can achieve impressive tracking performance. The challenges of this approach and techniques to alleviate the various problems are described. Experiments demonstrate that the proposed system can achieve an average tracking error of 0.89 m. Issues to be investigated further are discussed, which will enable the system to be deployed in places such as museums and shopping malls.

Positioning with a small number of anchors is an important issue in the study of acoustic indoor positioning systems (AIPS). We present PT-Sync, a novel approach to time-synchronize between a commercial off-the-shelf (COTS) speaker and a mobile device by leveraging acoustic sensing and reflected signals from the floor. PT-Sync enables ranging with one speaker and 2-D positioning with two speakers without any additional hardware. Our proposed time synchronization method requires height information from the floor in the calculation process, which we estimate by active acoustic sensing. Using a unique averaging technique and IMU enables robust height estimation. PT-Sync can be used in a variety of indoor environments and can be time-synchronized with already installed speakers. The results of the evaluation experiment confirmed that a synchronization error of 0.16 ms was achieved even at a distance of 6 m from the speaker, and that synchronization on the order of microseconds could be achieved. Furthermore, pedestrian tracking experiments confirmed that positioning of less than 38.9 cm can be achieved at the 90th percentile.

In this paper, we describe NL-Beep, which is a novel ranging method between multiple smartphones using acoustic sensing in non-line-of-sight (NLOS) environments. We need to know the distance between people to maintain social distancing because of the COVID-19 pandemic. Acoustic-based ranging methods that use a smartphone built-in speaker and microphone represent one approach to high-accuracy ranging methods. These methods can be used in line-of-sight (LOS) environments where there are no obstacles between the smartphones. However, in daily life, there are usually many obstacles, including pedestrians, and NLOS environments are the norm rather than the exception. Therefore, we propose the NL-Beep system for NLOS environments. To measure the distance between two smartphones without using a direct signal, NL-Beep efficiently uses a reflected signal from the ceiling in an indoor environment. It can also detect LOS and NLOS environments based on the profile of the received signal, and adapt the method used to estimate the distance between the smartphones. In our experiments, we used two smartphones in a room and estimated the distance between them for several placement configurations. In our results, we obtained 90th-percentile errors of less than 11.97 cm for the distances between the smartphones. In this paper, we also mention the effectiveness and limitations of NL-Beep.

In this paper, we propose a 2D localization method using a single speaker and a microphone. Our proposed method estimates the direction of the microphone based on the knowledge that the spectral amplitude of a speaker in each direction is different. Additionally, the distance between the speaker and microphone is estimated based on the received amplitude. The 2D location of the microphone is calculated from the estimated direction and distance values. We conducted real-environment experiments for two different types of speakers and confirmed that the 90-percentile localization error was 0.265 m. We also discuss the systematic errors in the distance estimation that occurred in the experiment, revealing that these errors were caused by the characteristics of the speakers.

This paper proposes a spot-communication method that can direct the spot to the user location adaptively. In the proposed method, the first transmitted signals are used to estimate user location, which is then reflected in the second transmitted signals; therefore, the spot directed to the user can be generated instantaneously. As the signals consist of mutually orthogonal sinusoidal waves, multiple spots can be generated simultaneously, and the shape of the spots can be controlled by overlapping them. Through real environmental experiments using four speakers, we confirmed that our proposed method can generate adaptive spots. Moreover, in the case that appropriate spots were not generated, comparisons were carried out with respect to the theoretical values to determine the cause.

The following topics are dealt with: indoor radio; indoor navigation; wireless LAN; smart phones; learning (artificial intelligence); pedestrians; Global Positioning System; radionavigation; mobile computing; Kalman filters.

In this paper, we describe a novel position-recognition method that uses passive acoustic signals from two previously installed speakers (passive acoustic sensing) and active acoustic signals from a smartphone's loudspeakers (active acoustic sensing). In passive acoustic sensing, a locus of positions for the smartphone can be calculated from the measured time difference of arrival from the two installed speakers. In active acoustic sensing, a chirp signal is transmitted from the speakers of the smartphone, and the distance to the side wall is measured from the propagation time of arrival at its microphone. We can obtain the smartphone position from our proposed model equations by combining these two results. In our experiments, we installed speakers at intervals of 10 m along a corridor and estimated the smartphone position at several places. From these results, we obtained 90th percentile errors of less than 0.224 m for 2-D positioning. We found that multipaths from the side wall were causing the positioning error in passive acoustic sensing, and the variance of the positioning error using the top microphone which was omnidirectional was smaller than the bottom one. When we introduced a weighting based on the result of the active acoustic sensing and the difference in the performance between microphones, the 90th percentile errors were reduced to less than 0.134 m.

This paper describes a spot-controllable data-transfer method. The proposed method generates a beam-shaped spot using two commercial off-the-shelf speakers. In our method, a symbol consists of a pair of sinusoidal waves having different angular frequencies. The width and direction of a beam-shaped spot are controlled by the angular-frequency difference between the sinusoidal waves and the transmission-time difference between the two speakers. Multiple spots can be generated by transmitting multiple pairs of sinusoidal waves based on the principle of orthogonal frequency-division multiplexing. By over-lapping multiple beam-shaped spots, the locations and sizes of the areas enabled to receive data are controllable. Experiments using four speakers and computer simulation show that the proposed method can generate controllable spots. An analysis of the errors in a real indoor environment indicate that they are caused by multipath signals, radiation damping of transmitted signals, and the incident/output angle characteristics of the microphone and speakers.

We describe an indoor localization technique for smart phones. Our new method, called the Frequency Division Multiplexing Phase Accordance Method (FDM-PAM), uses a beat called a sync pattern composed of a pair of sinusoidal waves with slightly different frequencies, which is similar to our original ultrasound ranging technique called the Phase Accordance Method (PAM). By generating multiple sync patterns with different central frequencies and transmitting them from different speakers, FDM-PAM conducts time-difference-of-arrival (TDOA) multilateration for localizing smart phones. In the current implementation of FDM-PAM, the 2D indoor position of a smart phone can be estimated. Three sync patterns are generated by using two out of six sinusoidal waves with frequencies ranging from 14.75 kHz to 17.25 kHz. The transmission of the sync pattern from the speakers lasts 4 ms. Through experiments, we have confirmed that FDM-PAM achieves accuracy of around 10 cm using only a short burst transmission, which indicates that the localization technique is sufficiently rapid and accurate.