SLAM
A core technology in robotics and autonomous driving that requires a device to simultaneously estimate its own position and construct a map of an unknown environment. Visual SLAM relies primarily on cameras as the main sensor.
Core Facts
SLAM (Simultaneous Localization and Mapping) technology is widely used in mobile devices such as robotic vacuum cleaners, enabling devices to build maps in real time and accurately locate themselves in unknown environments.
SLAM (Simultaneous Localization and Mapping) originated from the field of robotics, addressing the challenge of a device simultaneously localizing itself and constructing a map of an unknown environment.
Timeline (last 90 days)
Most novel-view synthesis methods assume that camera intrinsics and extrinsics are known, with poses pre-computed by SfM or SLAM; CoTL, however, directly takes pose-free multi-view images as input and needs to implicitly infer or absorb the geometric relationships between views.
SLAM is a technology that lets a device estimate its own pose while building a map in an unknown environment, and is the foundation for laser scanning and mobile mapping to obtain high-precision point clouds.
Cumulative localization drift in a SLAM system during a long-sequence walk causes the coordinate predictions of distant targets to shift further, creating an error stacking effect.
Traditional SLAM (such as ORB-SLAM) relies on handcrafted feature point matching; with the rise of deep learning, end-to-end approaches that jointly train depth estimation and pose optimization have emerged
Camera pose estimation typically relies on Visual Odometry or SLAM (Simultaneous Localization and Mapping) technology
Multi-user shared AR spaces rely on Spatial Anchor mechanisms, establishing a unified coordinate system across multiple devices via SLAM algorithms
Kinect was widely used in early computer vision and robotics research projects such as skeleton tracking, gesture recognition, 3D scanning, and robot SLAM navigation as a low-cost source of depth data.
Valerant introduces SLAM (Simultaneous Localization and Mapping) technology to reverse-engineer three-dimensional geometric structures and camera poses from continuous visual observations.
AMRs achieve flexible path planning and dynamic obstacle avoidance through SLAM, LiDAR, and visual perception, unlike traditional AGVs that rely on fixed magnetic strips or tracks.
SLAM algorithms are used to process raw sensor data to construct a geometric map of the environment, while semantic segmentation networks (such as SegFormer and Mask2Former) convert visual data into object category labels.
5 more timeline events
All Facts (17)
SLAM (Simultaneous Localization and Mapping) technology is widely used in mobile devices such as robotic vacuum cleaners, enabling devices to build maps in real time and accurately locate themselves in unknown environments.
75%VerifiedSLAM (Simultaneous Localization and Mapping) originated from the field of robotics, addressing the challenge of a device simultaneously localizing itself and constructing a map of an unknown environment.
65%Unverified商业级SLAM系统通常需要满足定位精度厘米级别、重定位成功率90%以上、在中低端移动设备上保持30fps以上实时帧率
75%UnverifiedMost novel-view synthesis methods assume that camera intrinsics and extrinsics are known, with poses pre-computed by SfM or SLAM; CoTL, however, directly takes pose-free multi-view images as input and needs to implicitly infer or absorb the geometric relationships between views.
50%UnverifiedSLAM is a technology that lets a device estimate its own pose while building a map in an unknown environment, and is the foundation for laser scanning and mobile mapping to obtain high-precision point clouds.
50%UnverifiedCumulative localization drift in a SLAM system during a long-sequence walk causes the coordinate predictions of distant targets to shift further, creating an error stacking effect.
50%UnverifiedCamera pose estimation typically relies on Visual Odometry or SLAM (Simultaneous Localization and Mapping) technology
50%UnverifiedTraditional SLAM (such as ORB-SLAM) relies on handcrafted feature point matching; with the rise of deep learning, end-to-end approaches that jointly train depth estimation and pose optimization have emerged
50%UnverifiedMulti-user shared AR spaces rely on Spatial Anchor mechanisms, establishing a unified coordinate system across multiple devices via SLAM algorithms
50%UnverifiedKinect was widely used in early computer vision and robotics research projects such as skeleton tracking, gesture recognition, 3D scanning, and robot SLAM navigation as a low-cost source of depth data.
50%UnverifiedValerant introduces SLAM (Simultaneous Localization and Mapping) technology to reverse-engineer three-dimensional geometric structures and camera poses from continuous visual observations.
50%UnverifiedAMRs achieve flexible path planning and dynamic obstacle avoidance through SLAM, LiDAR, and visual perception, unlike traditional AGVs that rely on fixed magnetic strips or tracks.
50%UnverifiedSLAM algorithms are used to process raw sensor data to construct a geometric map of the environment, while semantic segmentation networks (such as SegFormer and Mask2Former) convert visual data into object category labels.
50%UnverifiedPerception frameworks typically employ multimodal fusion architectures, including RGB-D cameras, LiDAR, and IMU devices, constructing geometric environment maps via SLAM algorithms.
50%UnverifiedSLAM algorithms are mainly divided into filter-based methods (such as EKF-SLAM and FastSLAM) and graph optimization-based methods (such as Cartographer's submap matching and sparse pose graph optimization).
50%UnverifiedModern SLAM (Simultaneous Localization and Mapping) solutions typically fuse data from multiple sensors such as LiDAR, cameras, and IMUs, and solve the problem using probabilistic filtering or graph optimization algorithms.
50%Unverified当前具身智能的主要技术瓶颈包括动态场景中SLAM的可靠性不足、长时序任务规划的计算复杂度，以及机器人硬件在能量密度和灵巧操控方面的物理限制
50%