Report: Cheap Lidar Unfit for Autonomous Highway Driving

The new wave of lower-cost lidar sensors being developed for self-driving vehicles may not provide sufficient performance to allow for hands-free driving at highway speeds, reports the MIT Technology Review.

In its analysis, MIT Review compares the differences between Velodyne Lidar Inc.’s top-end $80,000 HDL-64E unit and the company’s new $8,000 Puck device. The higher-priced unit lidar, which many early prototype autonomous vehicles have used for testing, has a 64-laser array spread at 0.4° with a stated range of 120 meters (400 ft), while the less expensive system is rated at 100 meters with a much lower resolution from 16 laser beams spread at 2.0°.

The authors say the 64 sensors can accurately render the surrounding area from a distance, but that’s it more difficult to spot and identify the same objects with the 16-unit system until they are much closer. The analysis advocates a 200-meter baseline for highway driving, noting that it takes vehicles traveling at 70 mph about 100 meters to come to a complete stop.

Initial solid-state lidar sensors being developed by several companies promise to lower costs to $250 or less. But MIT found these sensors also lack the necessary fidelity to operate safely and reliably at highway speeds. Valeo SA’s SCALA lidar, for example, provides four lines of data—rather than 64 or 16—with an angular resolution of 0.08°, which limits it’s use to vehicle speeds under 40 mph.

Velodyne and startups such as Quebec-based LeddarTech already are working on second-generation solid-state systems. These units promise to improve performance to match and eventually surpass that of current lidar.