Spatio-Temporal Calibration of a Vehicle-Mounted Microphone Array for Autonomous Driving
Organizations: Professorship of Autonomous Vehicle Systems and R. Taupitz with the Institute of Automotive Technology, TUM School of Engineering and Design, Technical University of Munich, 85748 Garching, Germany; Munich Institute of Robotics and Machine Intelligence (MIRMI).
Abstract
Microphones let an autonomous vehicle sense into areas its cameras and LiDAR cannot see, by picking up the sound that other traffic participants emit. Locating those sources and fusing them with the visual perception requires the microphone array to be calibrated and registered relative to the vehicle's other sensors. Existing calibration methods assume small, unobstructed indoor rigs, whereas a vehicle body casts an acoustic shadow that straight-line propagation cannot represent, and to our knowledge no calibration of an array distributed around a vehicle has been reported. We propose a pipeline that uses a balloon burst as a single cross-modal target, which the LiDAR locates in metric coordinates, every microphone hears as a sharp onset and a camera times. Each observation is weighted by its own timing variance, which discounts shadowed arrivals instead of discarding them and needs no model of the vehicle. On a full-size vehicle with eight microphones over a 4.9 m aperture, the held-out range residual of 6.37 cm approaches the 6.07 cm floor set by the onset timing noise, and the calibrated array tracks a moving source to within a few degrees of the vehicle's own LiDAR. A dozen bursts spread around the vehicle suffice, and spreading them matters more than adding more.
Figures & tables
| Work | Source (position from) | Aperture | Position error | Evaluation | |
| (a) Source localized by an independent second modality: the direct competitors | |||||
| 1 | Jacob and Haeb-Umbach [ 4 ] | Moving human talker (camera network, direction-of-arrival intersection) | Not reported (4 arrays in a m room) | 20 cm | Simulation only (image-method RIRs, modeled camera error) |
| 2 | Li et al. [ 5 ] | Six buzzers on a checkerboard (camera) | 0.69 m (0.4 m cube) | 2.4–6.1 cm | Real hardware, indoor |
| (b) Source positions estimated jointly with the geometry ( SLAM ) | |||||
| 3 | Zhang et al. [ 6 , 7 ] | Chirps from a robot-mounted loudspeaker (odometry, relative only) | Not reported (3 arrays indoors, m source trajectory) | 7.9 cm | Real hardware, indoor |
| 4 | Wang et al. [ 8 ] | Chirps from a robot-mounted loudspeaker (visual-inertial odometry, relative only) | Not reported (arrays spread over a m area) | 23 cm | Real hardware, indoor |
| RMS range residual [cm] | Microphone | |||||
| Variant | Held-out | vs. ref. | In-sample | displacement | Outl. | |
| LOS rays | SE | LOS | NLOS | [cm] | ||
| Full method (reference) | 6.37 | — | 6.37 | 25.14 | 0.00 | 13 |
| Observation weighting | ||||||
| Without the source-position term | 6.37 | 6.35 | 25.08 | 0.43 | 13 | |
| Without the SNR term | 7.75 | 8.49 | 20.83 | 8.14 | 8 | |