Occupancy Grid Mapping with 1D Probability Segments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating maps of physical environments around sensors in autonomous and ADAS-equipped vehicles are complex and computationally intensive due to the need for processing large amounts of data from sensors like lidars and radars to determine occupancy probabilities in occupancy grids.
Innovation Solution
A method that simplifies the calculation of occupancy probabilities by reducing the complexity by choosing one dimension of the sensor coordinate system based on measurement uncertainty, using a polar coordinate system where the distribution of probable values is represented as a segment, and approximating an arc of a circle as a straight line to simplify calculations, allowing for reduced computational effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the Inverse Sensor Model algorithm is used to calculate occupancy probabilities for all cells in the occupancy grid, then the accuracy of environmental mapping is improved, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent segments the two-dimensional occupancy probability calculation into two independent one-dimensional calculations: azimuth dimension and distance dimension. By dividing the complex 2D problem into simpler 1D sub-problems, the computational complexity is reduced while maintaining the accuracy of occupancy probability estimation for each cell in the occupancy grid.
Solution Approach 2:
The patent transforms the two-dimensional occupancy probability calculation into a product of two one-dimensional probability distributions. Instead of calculating P(occupancy) in 2D space directly, the method calculates P(azimuth) and P(distance) separately and combines them, effectively reducing the dimensional complexity of the computation while preserving the probabilistic accuracy.
2Reliability
If the Inverse Sensor Model processes large amounts of sensor data to determine occupancy probabilities, then the reliability of obstacle detection is improved, but the processing time exceeds real-time requirements
Solution Approach 1:
The patent segments the data processing workflow into independent azimuth processing and distance processing stages. Each stage processes one dimension of sensor data separately, allowing for optimized computation and reducing the overall processing time while maintaining reliable obstacle detection through the combination of both dimensional results.
Solution Approach 2:
The patent applies partial action by focusing computational resources on the most critical dimensions for obstacle detection. By processing azimuth and distance dimensions separately and combining results, the system achieves reliable detection without performing exhaustive processing of all possible cell combinations, thus reducing processing time while maintaining detection reliability.
3Loss of information
If the occupancy probability is calculated for every cell in the occupancy grid using the Inverse Sensor Model, then the completeness of environmental mapping is improved, but the computational resources required increase excessively
Solution Approach 1:
The patent segments the computational workload into independent one-dimensional probability calculations for azimuth and distance. This segmentation allows the system to maintain complete environmental mapping by evaluating all relevant cells in the occupancy grid, while significantly reducing computational energy consumption through the efficiency of one-dimensional probability distribution calculations compared to exhaustive two-dimensional processing.
Solution Approach 2:
The patent uses dimensionality change by transforming the computationally intensive two-dimensional occupancy probability calculation into a product of two one-dimensional probability distributions. This approach ensures complete environmental mapping coverage while reducing computational energy consumption, as one-dimensional probability calculations require fewer computational resources than direct two-dimensional calculations.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The method includes a step of measuring a potential position (µ) of an object, using a sensor; a step of identifying a segment (S), representing a distribution interval of probable values, associated with respective probability values, relative to the measured potential position, the segment extending along only one of the two dimensions of the sensor's coordinate system; and, the segment (S) penetrating potentially occupied cells (ci), a step of evaluating an occupancy probability of each potentially occupied cell (ci) comprising: - a step of determining characteristics of a portion of segment (xi), included in said potentially occupied cell (ci); - a step of calculating the occupancy probability of said potentially occupied cell as a function of the determined portion of segment, and using the probability density function.