Agent Trajectory Prediction Without HD Maps for Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face challenges in predicting agent behavior in areas lacking high-definition maps, requiring significant human and computing resources for map generation.
Innovation Solution
A method and apparatus that generate a feature map and occupancy flow map feature using sensor data, incorporating cross-attention and self-attention mechanisms to predict agent trajectories, and integrate text-based semantic information for enhanced prediction accuracy and reduced resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-definition maps are used for predicting agent behavior, then prediction accuracy is improved, but device complexity and resource requirements increase
Solution Approach 1:
The patent extracts and removes the dependency on high-definition maps from the prediction system. Instead of requiring external HD map data, the system generates prediction results using only sensor data from the autonomous vehicle, thereby reducing device complexity and resource requirements while maintaining prediction accuracy through alternative computational approaches
Solution Approach 2:
The system creates a virtual representation of the environment and agent behaviors through neural network models that process sensor data. This virtual copy replaces the need for physical HD maps, allowing the system to achieve accurate predictions through simulated environmental models rather than requiring actual map data
2Measurement precision
If high-definition maps are created for wide areas, then prediction accuracy is improved, but loss of time and human resources increase
Solution Approach 1:
The system performs preliminary processing of sensor data through neural network models to generate prediction results in real-time. By pre-training the models offline and performing inference online without requiring HD map generation, the system eliminates the time-consuming process of creating and updating wide-area maps while maintaining accurate predictions
Solution Approach 2:
The autonomous vehicle's system serves itself by generating prediction results using its own sensor data and onboard computational resources. This self-service approach eliminates the need for external HD map creation and maintenance processes, thereby reducing both time consumption and human resource requirements while maintaining prediction accuracy
3Measurement precision
If traditional prediction methods using past observation data and HD maps are used, then prediction accuracy is improved, but computing resource usage increases
Solution Approach 1:
The patent extracts and removes the HD map component from the traditional prediction pipeline, retaining only the essential prediction functionality. By using sensor data alone fed through optimized neural network models, the system achieves accurate predictions with reduced computing resource usage, eliminating the need to process and store large volumes of map data
Solution Approach 2:
The system changes the input parameters from traditional HD map data and historical observation sequences to optimized sensor data representations processed by neural networks. This parameter transformation enables the system to achieve comparable or superior prediction accuracy with significantly reduced computational complexity and energy consumption
Data Source
AI summary
A method for controlling autonomous driving of a vehicle is introduced. The method involves generating, based on an image obtained from a sensor of a mobility device, a feature map and a feature of an object, generating an occupancy flow map feature of the object, wherein the occupancy flow map feature comprises local information associated with a predicted trajectory of the object in the feature map, wherein the feature of the object is fused with the feature map, and wherein the predicted trajectory is estimated from the feature of the object fused with the feature map, outputting a predicted path of the object from a feature of the object fused with the occupancy flow map feature, wherein the predicted trajectory is determined, based on a reliability of the predicted trajectory, as the predicted path, and controlling, based on the predicted path, autonomous driving of the vehicle.