Lidar Conflict Resolver for Close-Object Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicle systems face challenges in accurately identifying and distinguishing between close objects in a driving environment using lidar point clouds, often misclassifying nearby objects as a single entity due to distance-based approaches, leading to errors in object recognition.
Innovation Solution
A lidar data segmentation system generates multiple segmentation scenes using different approaches, including raster-based and convex hull methods, and employs a conflict resolver to assign scores and select the most accurate scene, enabling the autonomous vehicle to differentiate between objects based on their geometrical characteristics and distances, even when they are closely spaced.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If distance-based segmentation approach is used, then segmentation process is simple, but object identification accuracy deteriorates when objects are closely spaced
Solution Approach 1:
The patent divides the segmentation task into multiple independent segmentation approaches (e.g., Euclidean clustering, voxel-based, density-based) that operate in parallel. Each approach generates its own segmentation scene, and the conflict resolver selects the most accurate scene based on scoring metrics. This resolves the contradiction by maintaining simple individual segmentation processes while achieving high accuracy through multi-scene comparison.
Solution Approach 2:
The conflict resolver acts as an intermediary between multiple segmentation approaches and the final object identification result. It receives segmentation scenes from different approaches, scores them based on geometric characteristics and consistency, and selects the best scene. This mediator component enables simple segmentation methods to achieve high accuracy without increasing the complexity of individual segmentation algorithms.
2Measurement precision
If multiple segmentation approaches are used, then object identification accuracy is improved, but system complexity increases
Solution Approach 1:
The patent merges multiple segmentation approaches into a unified system where each approach contributes its strengths. The conflict resolver combines the outputs by scoring and selecting the best segmentation scene, effectively merging the results without requiring complex integration logic in each individual approach. This maintains relative simplicity while improving accuracy.
Solution Approach 2:
Different segmentation approaches use different parameters and metrics (e.g., distance thresholds, voxel sizes, density parameters). The conflict resolver changes the evaluation parameters by scoring scenes based on geometric characteristics, consistency metrics, and object properties. This allows the system to adapt to different scenarios without fixing the system structure, managing complexity through parameter variation rather than structural complexity.
3Speed
If conventional distance thresholding is used, then processing speed is fast, but reliability deteriorates in varied driving environments
Solution Approach 1:
Instead of using a single distance threshold for all cases, the system applies multiple segmentation approaches with different threshold settings and parameters. Each approach performs partial segmentation with its own optimized parameters, and the conflict resolver selects the best result. This partial action approach maintains fast processing while improving reliability across varied environments.
Solution Approach 2:
The system dynamically selects which segmentation scene to use based on the current driving environment characteristics. The conflict resolver evaluates scenes using scoring metrics that adapt to environmental conditions, allowing the system to dynamically adjust its behavior without changing the underlying segmentation algorithms. This dynamic selection maintains fast processing speed while improving reliability.
Data Source
AI summary
An autonomous vehicle is described herein. The autonomous vehicle generates segmentation scenes based upon lidar data generated by a lidar sensor system of the autonomous vehicle. The lidar data includes points indicative of positions of objects in a driving environment of the autonomous vehicle. The segmentation scenes comprise regions that are indicative of the objects in the driving environment. The autonomous vehicle generates scores for each segmentation scene based upon characteristics of each segmentation scenes and selects a segmentation scene based upon the scores. The autonomous vehicle then operates based upon the segmentation scene.