Digital lidar technology offers many advantages, one of those advantages being sensor performance that can be dramatically improved through firmware updates. We have consistently delivered on that capability - shipping firmware over the past two years that included precision improvements, range improvements, and even thermal improvements for our digital lidar sensors. With Firmware 2.1.1we are continuing this trend, adding features that dramatically improve our sensors' perception performance, increase the detection probability of objects in the point cloud, and make our sensors fast, easy, and simple to work with.

Firmware 2.1.1 includes three major feature upgrades, along with bug fixes, to make our sensors even more effective for autonomy and smart infrastructure. With this new firmware release our customers will gain:

⇒ Improved reflectivity data quality

⇒ Improved detection probability of objects with Signal Multiplier mode

⇒ A new GUI for easily configuring sensors

You can read the detailed release notes here.

Improved Reflectivity Data Quality

Reflectivity data is a key input into many lidar perception algorithms. Reflectivity data from lidar sensors contains information about the inherent reflective property of objects that is independent of changes in the range of that object. To understand the significance of this, let's take a quick step back and review the types of data output by our sensor for every point in the point cloud:

Range: The distance of the point from the sensor origin, calculated by using the time of flight of the laser pulse

Signal: The strength of the light returned to the sensor for a given point. Signal for Ouster digital lidar sensors is expressed in the number of photons of light detected.

Near-IR: The strength of sunlight collected for a given point, also expressed in the quantity of photons detected that were not produced by the sensor's laser pulse

Reflectivity: The reflectivity of the surface (or object) that was detected by the sensor

Every point in the point cloud contains these four values, but what is apparent from these descriptions is that reflectivity is the only piece of data that contains information about the properties of the object, not light.

Okay, that's great, but why is it useful? Where signal varies with range (objects farther away return less light) and near-IR data varies with sunlight levels, reflectivity data is consistent across lighting conditions and range.

Let's take an example of a customer who is trying to consistently identify a 10% reflective target. If you depend upon the signal data, you will run into challenges as the signal data will be very high when the object is near to the sensor, but degrades exponentially as the target moves farther away from the sensor. With our improved reflectivity data, in contrast, that target will now return a consistent reflectivity value of ~10 across the entire range of the sensor.

In the two images below, notice the sign of the storefront. Prior to our improved reflectivity data you will see some data variation across the sign, represented as color changes in the point cloud (left image). With the improved reflectivity data, the sign is a consistent color indicating the consistency of the reflectivity values (right image).

This consistency of data for objects makes reflectivity data a powerful tool for perception algorithms, providing consistent data that can be used in object detection and classification.

We have achieved this improvement in reflectivity data through a combination of a new sensor calibration process, installed in January of this year, and this newly released Firmware 2.1.1, which takes advantage of our improved calibration.

For those taking advantage of this improved data we have a few details to share:

1. For OS0 and OS1 sensors that were purchased in February or later, customers will receive fully calibrated reflectivity data

2. For all OS2 customers and customers who have OS0s or OS1s purchased before February 2021, FW 2.1.1 delivers greatly improved reflectivity data

3. Reflectivity is most accurate for the OS1 and OS2 at ranges >10 meters, and for the OS0 at ranges >5 meters. Within those distances, reflectivity data quality is still improved but can be noisier

4. Reflectivity data will vary some across azimuth angles

Signal Multiplier Mode

With Signal Multiplier mode, our customers gain the ability to increase the signal strength (laser strength) of our sensors in certain modes. Our sensors, by default, operate in a Signal Multiplier mode of 1 (1x). Customers who operate the sensor with azimuth windows of 180° or less will now be able to double signal strength. Customers who only require an azimuth window of 120° or less are now able to triple their signal strength. The increased signal strength will improve the detection probability of all objects within the field-of-view.

What does increased detection probability mean for point cloud performance? It means that for dark objects that are not fully filled in with points, you will see more points on the target at all ranges. The ground rings of the point cloud will extend to a greater range (you can see this in the .gif below). Finally, increased signal strength will mean improved precision by ~20% in 2x mode and ~40% in 3x mode, to improve the quality of 3D maps and accuracy of localization algorithms.

Our customers using Signal Multiplier mode should be aware that the setting does have an impact on the maximum range of the sensor. By setting the Signal Multiplier mode to 2x or 3x you will reduce the maximum range at which the sensor will return points ('Maximum Representable Range' in the table below). For the OS0 there should be no visible impact on the point cloud data. For the OS1 and OS2 by setting the Signal Multiplier mode to 2 or 3 you will likely notice that the outer edges of the point cloud are slightly clipped.

Maximum Representable Range

Signal Multiplier Mode

OS0

OS1

OS2

1

240 m

240 m

480 m

2

135 m

135 m

240 m

3

90 m

90 m

160 m

New Configuration UI

Finally, with Firmware 2.1.1, we have introduced a new Configuration UI to quickly and easily configure the network, sensor mode, and synchronization timing settings of your Ouster digital lidar sensor.

To access the new Configuration UI, simply type 'https://os-XXXXXXXXXXXX' into your browser bar where XXXXXXXXXXXX is the twelve digit number on the top of your sensor. The number will start with either '99' or '12'. If your sensor is connected directly to your computer, add '.local' to the end of the serial number.

You are now directly connected to your sensor via the web browser, and within our 'Configuration' tab, can configure your sensor from the browser. You can update the networking settings, including changing the UDP destination IP address and UDP port numbers. You can adjust the sensor modes, including the new Signal Multiplier mode, and you can configure the timing settings in order to synchronize the sensor with other devices.

The Configuration UI is still in Beta mode, meaning that it is subject to change in the future, but should still work seamlessly today with all of your Ouster sensors.

Onward

This firmware release, like our past release, introduces significant upgrades to our lidar sensors. The significant improvements in reflectivity data will unlock new avenues in perception, enabling autonomy engineers to depend more and more on lidar as an object detection and classification tool.

With Signal Multiplier mode, we are giving even more control to our customers, empowering them to customize the sensor settings to better meet the exact needs of their deployment.

Finally, with the Configuration UI, our customers will see a significant improvement in the ease-of-use of our sensors, and more easily visualize all of the sensor settings available at their fingertips.

To download Firmware 2.1.1 and the accompanying documentation please visit our Downloads Center.

We look forward to delivering more significant improvements in Firmware 2.2 and beyond.

Attachments

  • Original document
  • Permalink

Disclaimer

Ouster Inc. published this content on 24 June 2021 and is solely responsible for the information contained therein. Distributed by Public, unedited and unaltered, on 24 June 2021 16:26:06 UTC.