Mobile LiDAR Trajectory Issues

I think most of us assume that the IMU – inertial measuring unit in a mobile LiDAR system only comes into play when GPS is not available, such as when you are in a tunnel or in an urban canyon. Turns out that is not the case. With the GPS updating 5 or maybe 10 times per second and the scanners spinning at 500,000 times per second in some cases the system is using the IMU all the time which introduces a lot more uncertainty into the solution.

Each IMU – derived position is sequential – it depends on the previous position which means any error in the starting position is going to be magnified as time progresses. This is what makes the kinematic use of LiDAR so much more challenging and why when you here people say that the accuracy of the point cloud is approaching the accuracy of the control that you want to be cautious.

To remove the uncertainty of the trajectory from the error budget will require very dense control which could defeat the purpose of using mobile in the first place.  It is important to remember every point in the cloud has its own unique georeferencing solution so if you are interested in 95% confidence in your results you have a lot of work to do.

This entry was posted in Mobile LiDAR, Uncertainty. Bookmark the permalink.

4 Responses to Mobile LiDAR Trajectory Issues

  1. Conor says:

    For sure – and what’s more – if you have a 100Hz INS and a 500KHz scanner, that leads to 5,000 points falling in between each INS update. So the vehicle position and orientation for 4,998 of those points is just being interpolated.

  2. Brian Bailey says:

    I do believe the overall message here is on point, and I have always said that the accuracy of the point is 100% dependant on the survey control network. In addition, every project is going to require different needs for control, and this is one of the largest variables in project budgeting. Also, it is why we must be cautious when organizations start publishing the “how to” control mobile mapping data papers and articles. I have been a part of many mobile mapping projects, and every one had different requirement in the control networks.
    With regards to the IMU, the above is a little misleading. The impact is VERY minimal in optimal GPS. A 500kHz LiDAR sensor isn’t spinning or scanning at 500kHz. The 500kHz is only the repition rate of the sensor, or how fast the laser is being fired. The scan rate (or mirror speed) on these sensors is 200Hz which is the same speed that the upper line IMUs are updating. In addition, of the 500,000 points/second that are being fired are being fired in a 360 degree field of view. This means that unless the vehicle is in a tunnel or going under a structure then 40% or more of the laser pulses being fired are being fired into the open sky and not actually being recorded as usable or recorded data.
    So, the variation would be in nanoseconds among point to point in a single revolution of the LiDAR sensor’s scanner. The drift from the best IMUs on the market don’t show significant drift until up to one minute of outage (as specified on their product information sheets). Also, drift isn’t constant. So, because it is gradual and based on times of outage, then the effects are very minimal if prevelant at all during the nanoseconds between IMU updates. Where the worries really come in is when the GPS isn’t updating and keeping the IMU in check, and this is why we find the need for more control around areas where the GPS was obstructed to the mobile mappers. We use the the survey control to make the corrections to the GPS outage and the drift that was associated in the trajectory from the IMU in the areas that the outages are experienced.

  3. Kin Yen says:

    The body orientation (yaw, pitch, and roll) accuracy has major effect of point accuracy as well. If you don’t know where your laser is point, it does not matter if you know exactly where you are. The uncertainty in the orientation will translate to larger error as the point are away from the vehicle trajectory. Yaw error are generally larger than pitch and roll, dual antenna enable yaw angle accuracy recovery faster after GNSS outage. Better IMU also give you better accuracy on orientation and thus give you better point accuracy when you don’t have any GNSS outage.

  4. Brian Bailey says:

    Kin, I respect your knowledge, experience, and opinion tremendously, but I have to respectfully disagree with you on the recovery of a dual antenna solution following GNSS outage. For example purposes, we will use the Applanix POSLV solution as an example since that seems to be the solution of choice for most mobile mapping systems out there. If you research the POSLV, you will find that at the upper end of the product line with regards to the IMU (500 level and up), the accuracies are the same (roll, pitch, yaw, x, y, z) with regards to 1 or 2 antenna. In fact, the highest level IMU (610) comes only configured in a 1 antenna solution. This is because the IMU is able to hold the yaw (heading) more precisely in GNSS outages. With regards to the recovery time, the single antenna solution is actually faster. In a 2 antennae solution, the GAMS (the distance measurement of the separation between the 2 antennae) are calculated during system calibration and dynamic driving prior to data acquisition. However, this measurement is constantly monitored and calculated real time by the solution. So, when a 2 receiver system losses GNSS, both receivers must reacquire satellite lock and then the system must start monitoring and calculating the separation measurement again before it will come back into a fixed integer. This takes additional time from the solution. However, with a single antenna, only one receiver must reacquire lock, and then the GNSS starts updating again at the selected intervals aiding in the heading measurements. Now, I agree that having 2 antennae are beneficial in certain applications or project environments, but these are often limited only to environments where the system is having to remain stopped with no movement for VERY extended periods of time. In addition, if a lower level IMU is selected (400 or 200 level), then a solution with 2 antennae is a must when trying to obtain a relatively high level of accuracy with the datasets. In contrast, with an upper level IMU, the 2 antennae may actually potentially have negative effects on the overall positional accuracy. When adding an additional GNSS receiver, it will only add additional GNSS noise to the trajectory. Many believe that adding an additional GNSS antenna will only add additional GNSS error, and since the GNSS data accounts for the largest source of error in the dataset, having 2 error sources will only increase the overall error with regards to positional accuracy. This additional error in the trajectory must be accounted for. I am not one who can argue against this logic, and I have found some validation to this. Again, I can only speak from my personal experience and having spent significant time with both types of solutions having 1 and 2 antennae, and the 1 antenna solutions with a higher grade of IMU (500 or 600 level) will actually recover faster, require less initial system calibration upon boot up, are more efficient during operations, produce better trajectories, and all of this is evident in the resulting datasets.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.