LIDAR ICP Alignment Trust Verification for Vehicle Trajectory Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately determining object alignments using the iterative closest point (ICP) algorithm, which can lead to incorrect tracking and trajectory planning if the alignments are not reliable.
Innovation Solution
A machine-learned model is trained to verify the accuracy of ICP algorithm alignments by comparing them to human-labeled data, determining whether the alignments should be trusted based on tolerance values that adjust with covariance, and controlling the vehicle accordingly, potentially using data from other sensors like radar if ICP data is unreliable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the ICP algorithm is used to determine correspondences between LIDAR data points, then object alignment and movement information can be obtained, but the alignment accuracy may be insufficient leading to incorrect tracking
Solution Approach 1:
A machine-learned model is introduced as an intermediary between the ICP algorithm and the object tracking system. The model verifies ICP alignment results by comparing them against learned patterns from training data, filtering out unreliable alignments before they affect tracking accuracy. This intermediary layer resolves the contradiction by maintaining the computational efficiency of ICP while adding accuracy verification.
Solution Approach 2:
The system implements feedback by using the machine-learned model to evaluate ICP alignment results and provide verification signals. The model learns from training data what constitutes reliable alignments and feeds this knowledge back into the tracking system, allowing the system to adjust its trust in ICP results based on learned criteria rather than blindly accepting ICP output.
2Productivity
If the ICP algorithm is run on every pair of consecutive point clouds, then object movement information is obtained, but computational complexity increases
Solution Approach 1:
Instead of always performing full ICP alignment verification, the system uses the machine-learned model to selectively verify alignments only when necessary. The model identifies cases where ICP results are likely unreliable and applies verification only in those scenarios, reducing overall computational complexity while maintaining tracking productivity.
3Ease of operation
If alignment data from ICP is used for trajectory planning, then driving decisions are made, but incorrect alignments lead to poor trajectory planning
Solution Approach 1:
The machine-learned model performs preliminary verification of ICP alignment results before they are used in trajectory planning. By pre-filtering and validating alignment data using learned criteria, the system ensures that only high-quality alignment information reaches the trajectory planning stage, preventing poor planning decisions while maintaining efficiency.
Data Source
AI summary
Aspects of the disclosure relate to training and using a model for verifying accuracy of ICP alignments or alignments between data points using an iterative closest point algorithm. For instance, a model may be trained using ICP alignment data, including alignments between an object appearing in LIDAR sensor frames. The training may also include setting a definition for a trusted ICP alignment. In this regard, the model may be trained such that, n response to receiving additional LIDAR sensor frames and corresponding additional ICP alignment data, output a value indicative of whether the additional ICP alignment data is trusted according to the definition. The model may then be used to control a vehicle in an autonomous driving mode by determining whether alignment data for object determined using the ICP algorithm should be trusted.