In the rapidly evolving world of robotics, autonomous vehicles, and surveying technology, SLAM (Simultaneous Localization and Mapping) has emerged as a foundational technology that enables machines to navigate and understand their surroundings. Whether it's a robot vacuum cleaning your home or a drone mapping a complex mine site, SLAM technology is working behind the scenes to understand where it is, what’s in the environment, and where to go next.
SLAM technology solves a chicken-and-egg problem that has long challenged robotics and autonomous systems: how can a device create a map of an unknown environment while simultaneously determining its own location within that environment? Traditional navigation requires either pre-existing maps or external reference points like GPS satellites. SLAM eliminates these dependencies, enabling devices to operate independently in GPS-denied environments such as underground mines, dense forests, or indoor spaces.
The technology works by using sensors - typically LiDAR (Light Detection and Ranging) or cameras to gather data about the motion of the system and the surrounding environment. Advanced algorithms then process this data in real-time, constructing a map while calculating the device's position and trajectory through space.
The SLAM technology landscape has diversified significantly in recent years, with various commercial and open-source solutions tailored to different applications and environments.
Emesent's SLAM algorithm - built from Wildcat and evolved to meet user needs for increased accuracy, represents the cutting edge of commercial SLAM technology, specifically designed for challenging underground and GPS-denied environments. The Wildcat algorithm was developed in a CSIRO research lab, and spun out into a startup enterprise - Emesent, where the co-founders Stefan Hrabar and Farid Kendoul continue to drive innovation. This system integrates sophisticated sensor fusion, combining LiDAR data with inertial measurements to achieve exceptional accuracy even in the most demanding conditions. Emesent SLAM excels in mining applications, tunnel surveying, and confined space mapping where traditional surveying methods prove impractical or dangerous. The system's robust algorithms can handle dynamic environments, dust, and poor visibility - conditions that would compromise less advanced SLAM solutions.
Beyond Emesent’s Wildcat SLAM, the SLAM ecosystem includes numerous alternatives. Beyond B2C options, commercial alternatives like GeoSLAM's ZEB series and NavVis systems cater to differing market segments, from heritage documentation to facility management. Each system presents trade-offs between accuracy, processing power requirements, cost, and suitability for specific environments.
It's important to distinguish between the LiDAR sensor itself — often called a "puck" due to its compact cylindrical shape — and the SLAM software that processes its data. The LiDAR puck is simply the hardware component that emits laser pulses and measures their return time to calculate distances, producing raw point cloud data. SLAM technology is the sophisticated software layer that transforms this raw data into actionable intelligence: building coherent maps, tracking position, correcting drift, and enabling autonomous navigation.
Multiple manufacturers often use identical or similar LiDAR pucks in their systems. For example, Velodyne or Ouster sensors appear across numerous SLAM platforms. However, two systems using the same LiDAR hardware can produce vastly different results depending on their SLAM algorithms. The distinction lies in how each manufacturer's software handles sensor fusion, processes the point cloud data, detects loop closures, manages accumulated error, and filters noise. Some manufacturers even use rotating LiDAR pucks to achieve greater coverage or redundancy. This means that when evaluating SLAM systems, the quality and sophistication of the software matters as much as — if not more than — the underlying sensor hardware.
Understanding SLAM accuracy requires examining how systems are evaluated and what factors influence their performance.
Unconstrained SLAM accuracy refers to how well the system performs without external reference points or corrections. This represents the "raw" performance of the SLAM algorithm as it operates independently. In unconstrained mode, a SLAM system might accumulate drift over long distances or extended operation periods—small positioning errors can compound over time, causing the map to gradually distort. High-quality systems like Emesent’s SLAM minimize this drift through advanced sensor fusion and loop closure detection, which recognizes when the device has returned to a previously mapped location and corrects accumulated errors.
Unconstrained accuracy is crucial for applications where no ground truth data exists, such as exploring previously unmapped caves or emergency response in disaster zones. Modern commercial SLAM systems can achieve unconstrained accuracies ranging from a centimetre to several meters, depending on the environment and distance travelled.
Constrained SLAM accuracy represents performance when the system can reference known control points or integrate with external positioning data. By incorporating surveyed ground control points or RTK data (when available) constrained SLAM dramatically improves accuracy. For Emesent’s SLAM, this approach integrates the RTK and ground control points as additional constraints during SLAM processing, eliminating cumulative drift and providing survey-grade results. In professional surveying applications, constrained SLAM can achieve <10mm 3D accuracy, making it suitable for engineering design and construction projects.
Environmental factors significantly influence SLAM accuracy and reliability, requiring operators to understand how conditions affect system performance.
Indoor spaces are generally easy for SLAM systems and can be measured highly accurately, though with a few potential challenges. Structured environments with walls, doorways, and furniture provide abundant geometric features that LiDAR-based SLAM can easily identify and track, increasing the SLAM’s position estimation and therefore accuracy. However, passing through tight indoor areas quickly, transitions such as doorways, or ‘repeated environments' such as long narrow hallways can be more challenging, causing SLAM to lose its place. Reflective surfaces such as glass walls or mirrors may also introduce false readings.
Underground environments — mines, tunnels, caves — represent some of the most challenging conditions for SLAM technology, which is precisely why systems like Emesent SLAM were developed specifically for these applications. Dust, moisture, and poor visibility test sensor performance, while repetitive tunnel structures lacking distinctive features challenge loop closure algorithms. The absence of GPS requires complete reliance on the SLAM system's autonomous capabilities. Despite these challenges, advanced SLAM systems achieve remarkable accuracy underground, typically maintaining centimeter-level precision through sophisticated sensor fusion and robust algorithmic approaches.
How do you make SLAM that works anywhere?
Outdoor environments are challenging for most SLAM systems. There are substantially fewer features for SLAM to identify and track, making achieving high accuracy results challenging. Open-sky environments allow integration with RTK for constrained accuracy and automatic georeferencing, which can enhance performance if the algorithm correctly optimises both the SLAM and RTK data. Introducing speed into the equation - for example when conducting a vehicle scan, can make it even harder for SLAM to track features in the environment accurately. Outdoor environments are a particular speciality of Emesent SLAM, as it was developed for the challenging, unknown, inaccessible environments that other SLAM systems struggle with.
The presence of moving objects fundamentally affects SLAM performance. Static environments allow systems to build consistent, stable maps, while dynamic environments with pedestrians, vehicles, or machinery require algorithms that can distinguish permanent features from temporary obstacles. Advanced SLAM systems incorporate object filtering and dynamic tracking to handle these situations, though accuracy may decrease slightly in highly dynamic settings. Generally, SLAM works well with dynamic environments as long as it can see and track enough static objects to understand the environment. Situations to be wary of - such as where SLAM may be focused on a moving truck, while on a vehicle mount, SLAM can track this feature and get confused. Similarly, elevators are SLAM’s kryptonite and users should take this into consideration.
Best Practices Guide for Scanning in Challenging Environments
Understanding Drift and Slip in SLAM
Drift is the gradual accumulation of small positioning errors over time as a SLAM system operates without external reference points. Imagine walking through a building while counting your steps - each slight miscalculation compounds, and by the time you've walked 100 meters, you might think you're several meters away from your actual position. In SLAM systems, this happens as tiny sensor measurement errors accumulate during processing, causing the map to slowly distort from its true geometry. High-quality SLAM algorithms minimize drift through loop closure detection (recognizing when you've returned to a previously mapped area and correcting accumulated errors) and sophisticated sensor fusion techniques. However, in large open spaces or long environments without distinctive features, some drift is inevitable in unconstrained SLAM operations.
Slip presents a different challenge - it occurs when the SLAM system temporarily loses tracking of its position, usually due to rapid motion, complete lack of environmental features, or sensor limitations. While drift is gradual, slip is sudden: the system essentially "skips" and must reorient itself, which can create discontinuities or jumps in the resulting map. Slip commonly happens during quick movements through featureless corridors, when scanning too fast in sparse outdoor environments, or when LiDAR data is temporarily compromised by dust or fog. To minimize both drift and slip, operators should maintain steady, controlled movement speeds, ensure the sensor captures sufficient environmental detail, and plan trajectories that create natural loops when possible.
SLAM technology has revolutionized how autonomous systems perceive and navigate their environments. From the specialized capabilities of Emesent's SLAM to the diverse ecosystem of alternative solutions, modern SLAM systems offer unprecedented flexibility and performance across varied applications. Understanding the distinction between constrained and unconstrained accuracy helps users select appropriate workflows for their precision requirements, while recognizing environmental influences enables better planning and more reliable results. As the technology continues advancing, SLAM will undoubtedly expand into new applications, making autonomous navigation increasingly accessible and accurate across all environments.
Want to see the results Emesent SLAM delivers? Accuracy, speed, versatility - it’s all here.
Article by: