LiDAR systems in applications such as autonomous mobile robots, drones, vehicles, and other commercial applications that demand compact, low-cost, and dynamic scanning will inevitably turn to MEMS mirrors as the beam-steering component. Beam scanning-based LiDAR architectures have a significant advantage as the full power and attention of the sensor is given sequentially to each point (voxel) in the scan. Competitive LiDAR designs typically utilize scanning and are differentiated by their scanning architecture and the specific hardware utilized, with the general goal of moving away from bulky mechanical and motor-based systems and toward compact silicon-based MEMS technology. Both single-axis and dual-axis MEMS mirrors are employed to enable two-dimensional (2D LiDAR) and three dimensional (3D LiDAR) point cloud sensing, respectively. The underlying time-of-flight sensor can be generic – a laser rangefinder or single-point LiDAR, with any typical wavelength or sensing method (pulsed ToF, AMCW, FMCW, etc.). The sensor is arranged with scanning elements which brings forth challenging trade-offs, discussed here. Architectures differ in whether transmitter and receiver are arranged coaxially or biaxially, each with its advantages and disadvantages. We present a hybrid architecture, Synchronized MEMS Pair LiDAR (SyMPL), which simplifies the coaxial design significantly and increases its efficiency by removing any beam splitting components or beam dumps. Multiple prototype LiDARs are compared and evaluated on the basis of SNR, scan speed, robustness to shock and vibration, eye safety, and resilience to mutual interference and echo signals. The work discusses the varying impacts on manufacturing and cost for applications demanding large volumes of LiDAR systems.
|
