Lidar has become a go to people measurement technology for Digital Mortar. Although lidar has numerous use cases including autonomous vehicles, mapping and robotics, in this review series we discuss the merits of different lidar sensor modes explicitly from a people (and vehicle) crowd measurement perspective.  We only review sensors that we have deployed on real world retail, airport, transit, stadium, casino, etc. settings.

Feel free to reach out to us at contact@digitalmortar.com with questions, feedback, to submit a sensor for review or if you would like to discuss what hardware might be right for your measurement needs.

Ouster OS0 Rev6

Ratings Summary

Overview

The Ouster OS0 is a puck style 360 mechanical LiDAR sensor intended for short range (30m or less) applications.

Beam Pattern (9/10)

Ouster OS0 has a 360 view and 35-degree field of view. By default, the beams are evenly distributed above and below horizon, but one of the sensor’s best people-measurement features is support for above horizon and below horizon beam pattern options.

In our deployments, sensors are mounted above human head height so any beams going upward are useless for people tracking. Tilting a sensor is one option, but as soon as you introduce any tilt the 360 benefits are lost, and the coverage area of each sensor is dramatically reduced.

The below horizon beam option allows sensors to be mounted right side up on CCTV mounts on columns/poles and interior/exterior walls and having the top beam run parallel to the ground and every other angled downward. The above horizon option allows sensors to be mounted inverted to ceilings or drop poles with similar effect. When possible, we particularly like the inverted option since it prevents any field of view loss due to the wall/column a sensor is mounted to and is aesthetically clean.

By back of the envelope math, a 64 channel OS0 sensor delivers 2.6 million points per second but the actual points that are potentially useful for people classification are as follows –

Evenly distrusted beam pattern mounted flat to wall – 40% (1.0m)

Evenly distributed beam pattern mounted flat to ceiling – 40% (1.0m)

Evenly distributed beam pattern mounted at an angle – 30% (0.8m)

Above/Below horizon beam pattern mounted flat to wall – 60% (1.6m)

Above/below horizon beam pattern mounted at an angle – 45% (1.2m)

Above/Below horizon beam pattern mounted flat to ceiling – 90% (2.3m

Beam Density (10/10)

The OS0 offers 32, 64 and 128 channel options. We love that the price difference between the three options, proportionate to the base sensor cost, is relatively minimal. That lets us recommend whatever is best for the use case. We find the 32-channel model to be a bit thin and typically gravitate toward the 64 or 128. The 128 is excellent in crowded environments and when not too many sensors are being deployed. The point cloud they generate is massive (5.2 million points per second) so they must be paired with a correspondingly powerful on-prem processor for handling the perception software.  The 64-channel model offers a good balance between beam density and scalability – they tend to be the option we use most frequently.

Coverage (8/10)

The OS0 sensors are purpose built for ‘short’ range with maximum tracking radius around 30m.  This may not be ideal for outdoor installations where you’re tracking people across a park or indoors in large open airport terminals, but it’s perfect for most indoor (especially retail) applications.  In most indoor environments, lidar beams are obstructed by walls, signs or displays prior to reaching their maximum 30m coverage radius capability, so the theoretical coverage range isn’t a limiting factor.

Value (10/10)

These are not the lowest cost sensors on the market, but they offer a lot more. In real-world deployments, you will love the custom beam configuration, the ability to step up to 64 or 128 channels, and the excellent people-measurement performance compared to lower cost sensor options. From a cost per square meter of coverage the OS0 Rev6 (with an above horizon beam patter and ceiling mounted) is likely the lowest cost sensor on the market.

The value is phenomenal.

 Power (4/5)

The OS0s are power hungry sensors. If direct powered to an outlet that’s not an issue, but we deploy most sensors with PoE power (like most LiDAR sensors the OS0s lack onboard PoE but can be PoE powered with the use of an in-line PoE splitter).  They draw up to 28W on startup and have a sustained power draw up to 20W. The draw on startup means that they need 60W PoE++, instead of the more common 30W PoE+, to ensure reliable startup 100% of the time. This is due to the expectation that some wattage is lost as you go down a cable run and from the PoE splitter – so if you start with 30W you might be closer to 23W by the time it gets to the sensor. If the sensor is asking for 28W and you’re supplying 23W, the sensor will be stuck, endlessly, in startup purgatory.

The need for PoE++ often means getting separate PoE injectors to power these sensors if your existing network switch doesn’t support PoE++.  In addition, the sustained draw of up to 20W sometimes means that people don’t have the power budget on a switch to support a large-scale sensor deployment and end up deploying PoE injectors anyway.  The power-hungry nature of the sensors is by no means an insurmountable obstacle (we choose to deal with in in a variety of deployment scenarios) but it’s a negative of the OS0 compared to an alternative LiDAR sensor that may use half the power and it you’re not careful, it can be a huge gotcha come installation time.

A Note About Version

Ouster currently ships Rev6 and Rev7 versions of the sensor with no intent to discontinue either line. This review is for the Rev6. Why not the latest version? Despite the somewhat misleading name, Rev6 and Rev7 are not like versions of the iPhone where newer is always better – they are essentially different products that target different applications. Rev7 sensors come in at a considerably higher price point and, though they deliver better accuracy in some situations, their advantages aren’t as relevant to full-journey people-measurement. It’s usually quite a bit better, in our use-cases, to deploy 2 Rev6 units (if necessary) than a single, comparably-priced, Rev7.