Simultaneous Localisation and Mapping

Simultaneous Localisation and Mapping (SLAM) is a core algorithmic technology enabling autonomous vehicles, Advanced Driver Assistance Systems (ADAS), and mobile robotics to achieve autonomous navigation. When operating in unfamiliar environments, a vehicle cannot rely solely on satellite navigation (GPS/GNSS) for centimetre-level precision, as signals are highly prone to drift or complete dropouts in high-rise urban canyons, tunnels, or basement car parks. SLAM allows a moving vehicle to perceive its surroundings in real time via on-board sensors (such as LiDAR, millimetre-wave radar, cameras, and Inertial Measurement Units [IMUs]), mapping its immediate vicinity in high definition while simultaneously computing its own precise relative coordinates within that map.

Core Technological Approaches and Operating Principles

SLAM systems are primarily categorised into two mainstream approaches: Visual SLAM (V-SLAM) and LiDAR SLAM.

Visual SLAM relies mainly on high-resolution on-board cameras to capture environmental textures, estimating motion trajectories through geometric triangulation and feature-point matching across successive frames. While cost-effective and rich in semantic data, it remains sensitive to ambient lighting changes. In contrast, LiDAR SLAM emits laser pulses via high-speed rotating units, using time-of-flight ranging to construct dense 3D "point cloud" spatial maps. This delivers exceptional ranging precision and all-weather operational capability. In advanced autonomous driving applications, a multi-sensor fusion approach is typically deployed, tightly integrating LiDAR, camera, and IMU data to ensure positioning redundancy and robust reliability across complex, dynamic driving environments.

Application Scenarios and Market Value

SLAM technology serves as the foundational cornerstone for high-level autonomous driving capabilities, including highway/urban navigation-assisted driving, Automated Parking Assist (APA), and Automated Valet Parking (AVP).

In multi-storey or basement car parks and other covered areas without satellite reception, vehicles cannot rely on conventional satellite navigation or pre-loaded HD maps. In such scenarios, the on-board SLAM system dynamically constructs a real-time map of pillars, walls, and vacant parking bays on the fly, guiding the vehicle safely around obstacles and into the intended parking bay. This equips intelligent vehicles with true autonomy and pinpoint navigation capability in "mapless" zones or satellite blind spots.

Safety Guidelines and Physical Limitations 

Drivers must never blindly rely on SLAM algorithms for autonomous navigation in severe weather conditions (such as torrential rain, dense fog, or heavy snow) or when sensor fields of view are heavily obstructed or soiled:

SLAM algorithms depend entirely on sensors capturing consistent environmental features. When camera lenses are obscured by road grime or LiDAR units are compromised by heavy snow and ice, the system risks losing tracking points or generating false matches, leading to positional drift or complete mapping failure. Drivers must remain alert at all times and be prepared to resume manual control immediately whenever a sensor obstruction warning is issued.

Never overlook the limitations of SLAM algorithms in filtering moving targets in highly dynamic environments (such as busy car parks with heavy pedestrian traffic and crossing vehicles): 

Conventional SLAM algorithms fundamentally assume a static environment. In settings with numerous moving objects, the algorithm can easily misinterpret dynamic targets as stationary landmarks, causing severe distortion in trajectory calculation. Advanced autonomous driving systems must incorporate deep learning and dynamic object filtering mechanisms to guarantee positioning safety in complex environments with mixed pedestrian and vehicular traffic.