LiDAR Soft Consistency Checking for Non-Impeding Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems, particularly in autonomous vehicles, often misclassify non-impeding objects such as fine particulate matter as solid objects, leading to hazardous vehicle trajectories.
Innovation Solution
A method using lidar systems to emit and receive pulses, combined with machine-learned models and voxelization techniques, to determine the probability that an object is non-impeding by analyzing lidar intensity, multiple return pulses, and motion status, and applying Bayesian probabilistic fusion to ensure accurate classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If lidar systems detect objects by measuring reflected pulses, then object detection capability is improved, but false positive indications of solid surfaces are generated by non-impeding objects like fog, smoke, dust, and steam
Solution Approach 1:
The system performs multiple return pulse detections and uses the consistency of measurements across these detections as feedback to verify object solidity. Non-impeding objects produce inconsistent measurements across multiple pulses, while solid objects produce consistent measurements, allowing the system to distinguish between them and reduce false positives
Solution Approach 2:
The system performs more detections than the minimum single pulse by utilizing multiple return pulses and aggregating measurements. This excessive action of multiple detections allows statistical analysis of measurement consistency, improving reliability by filtering out transient reflections from non-impeding objects
2Reliability
If machine-learned models classify detected objects as non-impeding, then vehicle trajectory safety is improved, but computational resources and processing time are increased
Solution Approach 1:
The system pre-computes voxelizations of the environment and pre-processes lidar point data into structured formats before object classification is needed. This preliminary organization of data accelerates the machine-learned model's processing time during critical trajectory planning moments
Solution Approach 2:
The system divides the environment into discrete voxels and processes lidar returns by voxel, segmenting the computational task into manageable units. This segmentation allows parallel processing and optimizes the efficiency of the machine-learned model by processing spatially localized data rather than entire point clouds
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
Improves the accuracy of object detection by reducing false positives, enabling safer and more efficient vehicle trajectories by distinguishing between solid and non-impeding objects, thereby enhancing safety and reducing computational resources.
Implementation Method 1
autonomous vehicles may be configured with lidar systems that use lasers to emit pulses into an environment and sensors to detect pulses that are reflected back from surfaces of objects in the environment
Implementation Method 2
sensors to detect pulses that are reflected back from surfaces of objects
Implementation Method 3
determining one or more voxels associated with the environment; determining one or more individual sensor data points of the set of the sensor data associated with individual voxels
Data Source
AI summary
Techniques for determining a probability that an object that has been appropriately labeled as a non-impeding object are disclosed. Non-impeding objects may be objects in an environment that do not substantially affect the operation of a vehicle in the environment, such as steam, fog, smoke, etc. A machine-learned model may classify such objects as non-impeding objects. The designation may be checked by using the lidar data and measurements associated with the lidar points associated with the object to determine a probability that the object is a non-impeding object. The lidar data may be aggregated within the individual voxels of a voxelized representation of the object for use in the probability determination.