SLAM (Simultaneous Localisation and Mapping) is poised to be the next-big-thing in surveying. It will disrupt how surveyors generate maps, engage with data, and deploy reality capture technology within an industrial context.
There are many different types of SLAM (and SLAM scanners), and it’s now possible to rapidly capture data avoiding complex or technical processes. But the transition from static to mobile scanning isn't as simple as buying a SLAM scanner. SLAM needs to be used in the right way at the right time to be effective.
Here, we’ve captured the rough-and-ready overview — what SLAM is really best placed to deliver in the reality capture context.
SLAM is attempting to calculate a location within a map while generating that map at the same time. This requires some shortcuts and approximations. SLAM algorithms are not aimed at perfection, only operational effectiveness.
In the context of wearable technology, such compromises are essential to keep scanners small and somewhat affordable. To implement SLAM, there has to be the usual trade-off between speed, processing power and accuracy. For example, most portable SLAM scanners are currently in the 3cm accuracy range, which limits their application.
A traditional strategy for preventing propagation error within this context is to place stationary targets throughout a scene (with known and measured locations) and using that to ground the mobile scanning data. However, you can improve the accuracy of SLAM even more by overlaying registered mobile scans with point clouds generated by traditional laser scanners.
For example, if you are using SLAM on a construction site to check for changes against planning, you may already have a higher-accuracy scan. The right software will allow you to cross-reference key features within those scans and use millions of points of reference (rather than just a few dozen) to align that SLAM data.
Pro tip: This strategy doesn’t actually change the inherent precision of the scan, but it does help minimise propagation of error, and can improve the accuracy and precision of your end output.
SLAM algorithms will improve over time. It's important to watch out for these developments. But, fundamentally, SLAM scans are approximations — and that is what they are meant to be.
Surveyors don't need to "solve" the precision problem. Layering scan can help improve precision. However, the real takeaway is actually the importance of using SLAM for the right job. That means looking at ways to integrate SLAM and mobile scan data into your broader reality capture and modelling data sets.
Related to the problem of accuracy is the complex nature of SLAM processing. The more data you have, and the more accurate your outputs need to be, the harder it is to undertake that processing at speed. This is why almost all SLAM algorithms need a “rough solve” in the field and then a complex loop closure (and final SLAM solve) offline, later.
When you consider the kind of small device needed to move through complex environments such as a warehouse, tunnel, or cityscape, there is not enough on-board processing capacity to handle the complex calculations required.
In these and a few other circumstances, the speed of computation is a huge problem. This is particularly true when looking at fully autonomous robotics — which needs to process SLAM scans in near real-time. However, for surveyors, it's not that big an issue.
There is potential to offload data processing to the cloud to take advantage of its real-time parallel computation. If coupled with more cloud-oriented algorithms, this generates significant performance gains.
By using a combination of cloud-based processing, 5G connectivity and better algorithms, surveyors can manage even massive data sets more easily. Again, this is the key to flexibly using mobile scanning data sets with wider reality capture data sets. These same tools have already transformed static point cloud processing and will be essential to being able to use mobile data within your workflows.
Suggested reading: Are You Ready For The Cloud? A surveyor’s guide to the future of 3D laser scanning
More frequent surveys are becoming the norm. The development of Level 3 BIM (Building Information Management) will drive the need for more live verification and progress reports. 3D LiDAR-based SLAM can deliver such "on-demand" scans, scaling with the increasing Level of Detail (LOD) needed for each stage.
One application which really lends itself to SLAM is Scan-to-BIM — comparing as-built against actual designs. For the built environment, this opens immense opportunities for fast, accurate 3D models created in the minimum amount of time:
The enhancement in data quality, due to improvements in layering mobile and static scans, will mean that surveyors can use mobile devices for even more BIM applications. Start looking for SLAM capability in your mobile indoor mapping scanner procurement and your registration software.
A large part of the challenge of SLAM is the move to mobile scanners. Inevitably, the number of points to process increases dramatically. With the increasing capability of SLAM-based laser scanners, it won't be uncommon for hundreds or even thousands of scans to be captured daily.
The simplicity and speed of mobile scanning make it easy to take more scans over time and cover greater distances. This is a strength of the technology, but it's also a weakness.
The size of mobile datasets not only causes processing challenges, it makes it essential to piece together already registered scans. However, this is not a capability most SLAM registration software was built to achieve.
Being able to maintain details on how your scans link together will be vital; the days of making sketch notes of site networks will be long gone. More than likely, these paper records will get lost, damaged or will be out of date by the time they get back to the office.
Ensure that you are using mobile technology to capture scan locations. Also, make sure that file names and naming conventions are clear when you are storing scans – especially if you are looking to overlay scans from different sensors and scanners. More fundamentally, you need to make sure you have the software available to use and overlay these scans once they are produced.
SLAM is inherently mobile. We have enough examples already in the wider world of how transformative making things mobile can be. For anyone involved in the acquisition of surveying data, SLAM is a game-changer.
By removing laborious set-ups from the equation, time and cost savings are potentially enormous. SLAM based mobile mapping systems can be over ten times faster at acquiring data than traditional methods.
The problem with SLAM is accuracy and precision. However, this is only a problem if you are looking at it as a direct replacement for static laser scanners. A wide range of use cases will fit the "just-good-enough quality" approach of SLAM.
For example, a total station is far more accurate than a 3D laser scanner — but this doesn’t undermine the value of laser scanning. In fact, these tools are commonly used in complementary ways. SLAM is just one more option to integrate into your process. Layering these scans will improve their quality. But making sure that you use SLAM for the right job is the most important factor. You want to incorporate all of your data into higher detailed maps for a dynamic, frequently updated single-source-of-truth.
Fundamentally, all this will be made faster and easier with automated cloud-based software tools to collect, register, and ensure the quality of these scans. Scanning is going mobile and surveyors need to update.