Vehicle Environment Patterning With Attention-Based Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning algorithms for autonomous driving, such as those using recurrent units, struggle to explicitly differentiate between distinct objects in a vehicle's environment and incorporate legacy information to improve dynamic property estimation and object tracking.
Innovation Solution
A computer-implemented method employing an attention algorithm to generate a joint spatial-temporal data set by explicitly correlating current and memory data, allowing for improved object tracking and dynamic property estimation by considering legacy information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If recurrent units are used to derive spatial and temporal patterns from sequential sensor scans, then the model can process sequential data and generate environment patterns, but the model cannot explicitly differentiate between distinct objects within a receptive field and cannot effectively incorporate legacy information to improve dynamic property estimation
Solution Approach 1:
The patent segments the processing of sequential data by introducing distinct components: a memory module that stores legacy information from previous time steps, and an attention mechanism that selectively attends to relevant objects and time steps. This segmentation allows the model to differentiate between distinct objects while preserving and utilizing legacy information, resolving the contradiction between object differentiation precision and legacy information utilization.
2Reliability
If machine learning algorithms aggregate sensor data over previous points in time to generate memory state, then temporal patterns can be derived, but the algorithms cannot explicitly correlate sequentially recorded data to improve velocity estimation and object tracking
Solution Approach 1:
The patent introduces an attention mechanism as an intermediary between the memory state (aggregated historical data) and the current sensor scan. This attention mechanism selectively weights and correlates data from previous time steps with current observations, enabling explicit correlation of sequentially recorded data to improve velocity estimation and object tracking reliability without requiring complex manual correlation logic.
3Measurement precision
If the model uses implicit correlation of features in memory state and current data, then spatial-temporal patterns can be generated, but the model cannot explicitly differentiate objects or incorporate legacy dynamics information for improved velocity estimation
Solution Approach 1:
The patent implements feedback through the attention mechanism, which uses the current sensor scan and memory state to compute attention weights that determine how much legacy information from previous time steps should be incorporated into the current estimation. This feedback loop enables the model to explicitly differentiate objects and incorporate legacy dynamics information for improved velocity estimation, while the automated attention computation maintains ease of operation without requiring manual intervention.
Data Source
AI summary
A method is provided for determining patterns related to an environment of a host vehicle from sequentially recorded data. Respective sets of characteristics detected by a perception system in the environment of the host vehicle are determined for a current point in time and for a predefined number of previous points in time. A set of current input data associated with the sets of characteristics is generated for the current point in time, and a set of memory data is generated by aggregating the sets of characteristics for the previous points in time. An attention algorithm is applied to the set of current input data and to the set of memory data to generate a joined spatial-temporal data set, and at least one pattern is determined for the environment of the host vehicle from the joined spatial-temporal data set.