3D Sub-Grid LiDAR Pose Estimation for Low-Compute Robots
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for robot pose estimation in 3D space, such as scan matching and grid map-based Monte Carlo localization, face high computational loads and are prone to errors due to limited point cloud data or environmental changes.
Innovation Solution
A 3D sub-grid map-based robot pose estimation method that reduces computational requirements by performing calculations within a 3D sub-grid centered around the robot, using LiDAR scan data to generate and transform particles, and filtering them based on similarity and obstacle probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If scan matching-based algorithms are used for robot pose estimation, then positioning information can be obtained by matching 3D coordinates, but the computational load becomes high and incorrect positions are predicted when point cloud data is insufficient or environment changes significantly
Solution Approach 1:
The patent divides the computational space into discrete 3D grid cells, where each cell represents a possible robot pose. Instead of continuously comparing point clouds across the entire workspace, the algorithm segments the search space into manageable grid units, allowing efficient computation of pose probabilities for each cell independently.
Solution Approach 2:
The patent transitions from traditional 2D grid map localization to 3D grid map localization by adding the vertical dimension. This enables robots to perform pose estimation in three-dimensional space, accommodating applications like drone navigation and multi-level warehouse operations where vertical position is critical.
2Adaptability or versatility
If grid map-based Monte Carlo localization algorithm is used in 3D space, then position estimation in 3D space can be achieved, but computational requirements and memory usage increase significantly
Solution Approach 1:
The patent segments the 3D workspace into a grid map where each cell stores occupancy probability and pose information. This segmentation allows the system to handle 3D localization by processing discrete grid cells rather than continuous space, reducing computational complexity while maintaining 3D navigation capability.
Solution Approach 2:
The patent changes the representation parameters from traditional particle filters to grid-based occupancy probabilities. By representing the environment as a 3D grid with probability values rather than using numerous particles, the system reduces memory usage and computational requirements while maintaining accuracy in 3D pose estimation.
3Speed
If traditional Monte Carlo localization is applied to 3D space, then real-time position estimation can be performed, but expensive memory and CPU resources are required
Solution Approach 1:
The patent creates a simplified 3D grid map representation that copies essential spatial information from the environment. This grid-based copy allows the system to perform rapid pose estimation by comparing sensor data against the pre-built grid structure, achieving real-time performance with reduced resource requirements compared to full 3D point cloud processing.
Solution Approach 2:
The patent changes the state representation from continuous particle coordinates to discrete grid cell indices and probability values. This parameter transformation enables more efficient storage and computation, allowing real-time 3D pose estimation on systems with limited memory and CPU resources.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method enables real-time pose estimation on low-performance personal computers by reducing computational costs and minimizing pose estimation errors caused by dynamic obstacles, while maintaining accuracy in 3D space.
Implementation Method 1
a light detection and ranging (LiDAR) part configured to emit light and detect reflected light from objects in a global space to generate and transmit LiDAR scan data
Data Source
AI summary
Embodiments relate to a method and a robot for estimating a pose, and the robot estimating the pose using a 3-dimensional (3D) sub-grid map includes a main body part, a transfer part configured to move the main body under the control of the main body part, and a light detection and ranging (LiDAR) part configured to emit light and detect reflected light from objects in a global space to generate and transmit LiDAR scan data to the main body part, wherein the main body part includes a personal computer (PC) estimating the position and orientation of the mobile robot, the PC including a LiDAR scan data acquisition module configured to acquire LiDAR scan data for each sub-grid of a 3D grid map based on the robot, a particle generation module configured to generate robot candidate particles on the global map, a LiDAR scan data transformation module configured to transform the LiDAR scan data acquired by the LiDAR scan data acquisition module based on the pose of the robot to the pose of particles generated by the particle generation module, and a sub-grid projection module configured to display the transformed LiDAR scan data from the LiDAR scan data transformation module onto the 3D sub-grid based on the robot.