Scene Encoding Apparatus Using Local Obstacle Tensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing trajectory prediction technologies face inefficiencies due to the need for frequent normalization of coordinate systems and recalculation of scene encodings for each time point, especially in streaming data scenarios where obstacle data and map data are constantly updated, leading to increased calculation time and reduced prediction efficiency.
Innovation Solution
A scene encoding generating apparatus that transforms obstacle and map data into a local coordinate system based on the position and movement state of each obstacle, generating a local obstacle tensor which is then input into a scene encoder to produce a scene encoding that can be used for trajectory predictions without requiring constant normalization of coordinate systems across different time points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If coordinate system normalization is performed frequently for each time point, then trajectory prediction accuracy is maintained, but calculation time increases and prediction efficiency decreases
Solution Approach 1:
The patent pre-establishes a local coordinate system for each obstacle based on its position and movement state. This preliminary coordinate system setup allows subsequent trajectory predictions to use the pre-computed local frames without repeatedly performing coordinate normalization, thereby reducing calculation time while maintaining prediction accuracy.
Solution Approach 2:
The patent transforms the global coordinate system into local coordinate systems for each obstacle, changing the reference frame parameter from global to local. This parameter change eliminates the need for frequent coordinate normalization operations while preserving the spatial relationships needed for accurate trajectory prediction.
2Measurement precision
If scene encoding is recalculated for each time point, then prediction accuracy is maintained, but productivity decreases
Solution Approach 1:
The patent performs preliminary transformation of obstacle data into local coordinate systems and generation of local obstacle tensors before the actual prediction process. This preliminary action consolidates the scene encoding calculation, allowing it to be performed once rather than repeatedly for each time point, thereby improving productivity while maintaining accuracy.
Solution Approach 2:
The patent segments the coordinate transformation process into individual local coordinate systems for each obstacle, rather than performing a single global coordinate normalization. This segmentation allows each obstacle's local frame to be independently established and reused, reducing the overall computational burden and improving prediction efficiency.
Data Source
AI summary
A scene encoding generating apparatus is configured to execute the following operations. The scene encoding generating apparatus generates a local coordinate system based on a position and a movement state corresponding to each of a plurality of obstacles in a time point. The scene encoding generating apparatus transforms the position and the movement state corresponding to each of the obstacles to the corresponding local coordinate system to generate a local position and a local movement state. The scene encoding generating apparatus generates an obstacle tensor corresponding to the obstacles based on the local positions and the local movement states corresponding to the obstacles, wherein the obstacle tensor is corresponding to the time point. The scene encoding generating apparatus inputs the obstacle tensor into a scene encoder to generate a scene encoding, wherein the scene encoding is configured to be inputted into a decoder to generate a trajectory prediction.