New 3D LiDAR applications have exploded in the last year, largely due to the release of Velodyne’s economical VLP-16 ‘Puck’ LiDAR sensor. Designed to be the ideal robotic eye, 3D LiDAR can turn any vehicle into an autonomous platform by allowing it “to see.” Boats, cars, drones, and even trains are being equipped with these sensors, which are capable of delivering 300,000 points of XYZ spatial data per second in 360 degrees. While this standard design has been suitable for most autonomous platforms, there are areas where the product could be optimized for better performance. One of the most common requests is for alterations to the firing angle of the 16 laser channels.
The VLP-16 has a 360 horizontal Field of View (FoV) with a +15 to -15 degree vertical FoV angle. Each of the 16 channels is spaced and directed about two degrees apart (see manual for specifics). The sensor is capable of operating at any orientation so engineers can tilt the sensor on its mount to achieve +5 to -25 degrees, 0 to 30 degrees, +10 to -20, etc. There is a limit to angle positions as the laser is fixed, so while design changes to the VLP-16 are possible, the related cost increases to create a highly customized sensor are prohibitive. However, one possible workaround to change the angle of the lasers beyond their range in a fixed position could involve a using a mirror.
The Mirror Visor is a concept I came up with while selling Velodyne sensors on the east coast. This is not a product, but rather an idea that other engineers can use to potentially customize their VLP-16 sensors. A rigid mirror can be attached to the top rim of the puck in order to reflect the lasers in a different direction. The law of reflection states that the angle of incidence and the angle of reflection both lie in the same plane and are equal. By keeping this information in mind, the mirror visor could be adjusted with an orthogonal angle or even a slightly downward angle. By changing the angle of the lasers, the Puck can strike different targets. This is particularly effective in automated guided vehicle (AGV) applications because it allows the robot to create a series of concentric lines around itself, almost like a curtain, that can scan for obstacles and objects.
That said, engineers should keep be aware of how a sensor interprets data that has been reflected off a mirror. The sensor does not inherently know that the laser angle has been changed along its firing path, so the data appears as if it is behind the mirror. This is not a sensing error, and a correction can be factored into the software to account for this adjustment because each distance data point from the VLP-16 is tagged with the associated vertical channel and rotational angle. So a programmer could specify that channel 16 is reflected 40 degrees downward and that an adjustment would be performed in the point cloud transform function.
Each LiDAR application is different and may require specific sensor hardware features. The mirror visor is a way to change the performance characteristics of the VLP-16 3D LiDAR sensor without having to completely redesign it. Please see the following images which show the proof of concept. This article, experiment, and design were created by Frank Bertini. You can contact him at [email protected]
An adjustable mirror is used to alter the beam travel of the lasers and reflective tape is used as a target.
The mirror creates a slightly angle, about 10-20 degrees
A more severe angle can reflect lasers at a sharper downward angle.
Notice that only the bottom 6 or 7 lasers are hitting the target. The bottom of the square target and the long skinny rectangle portion of the target are currently out of the FoV.
By reflecting the laser downward, more of the bottom of the square target can be seen. (Note: Our viewing program does not account for the mirror and the reflective target appears to be behind the mirror. This effect can be eliminated in software.) A portion of the long skinny rectangle can also be seen at the very top of the image. This feature was not visible in the before image without the mirror.