Behavior-Guided Path Planning for Vehicle-Agnostic Autonomous Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicle systems face challenges in scalability and efficiency due to the need for separate deep neural networks for each vehicle make and model, requiring significant computational resources and manual labeling for training, which limits their ability to adapt to different environments and vehicles.
Innovation Solution
The system uses behavior-guided path planning by automatically generating ground truth data from sensor information to train machine learning models, allowing them to predict paths in various environments without the need for manual labeling and enabling vehicle-agnostic control across different vehicle types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If separate deep neural networks are trained for each vehicle make and model, then vehicle-specific control accuracy is improved, but device complexity and computational resources required increase significantly
Solution Approach 1:
The patent applies universality by training a single deep neural network model that can be deployed across multiple vehicle makes and models. The system achieves this by collecting sensor data and ground truth trajectory data from diverse vehicle types during operation, then training one universal model on this aggregated dataset. This single model replaces the need for separate vehicle-specific models, reducing system complexity while maintaining adaptability through its ability to learn from heterogeneous data sources.
2Measurement precision
If manual labeling is used to train deep neural networks, then training data quality is improved, but time and effort required for data preparation increase significantly
Solution Approach 1:
The patent implements self-service by enabling the system to automatically generate its own training data through operational data collection. During vehicle operation, sensor data is collected along with ground truth trajectory information from GPS and other positioning systems. This automatically captured data serves as high-quality training labels without requiring manual annotation, allowing the system to continuously improve its model while eliminating time-consuming manual data preparation processes.
3Measurement precision
If end-to-end trained deep neural networks output vehicle-specific commands, then control precision for specific vehicle types is improved, but scalability to different vehicle makes and models deteriorates
Solution Approach 1:
The patent applies parameter changes by training the deep neural network on aggregated data from multiple vehicle makes and models, effectively changing the training parameters to include diverse vehicle characteristics. The system collects sensor data and corresponding ground truth trajectories from various vehicle types, then uses this heterogeneous dataset to train a single model that adapts to different vehicle parameters. This approach maintains control precision across vehicle types while achieving scalability, as the model learns to adjust its predictions based on the input data rather than being hard-coded for specific vehicle configurations.
Data Source
AI summary
In various examples, a machine learning model—such as a deep neural network (DNN)—may be trained to use image data and/or other sensor data as inputs to generate two-dimensional or three-dimensional trajectory points in world space, a vehicle orientation, and/or a vehicle state. For example, sensor data that represents orientation, steering information, and/or speed of a vehicle may be collected and used to automatically generate a trajectory for use as ground truth data for training the DNN. Once deployed, the trajectory points, the vehicle orientation, and/or the vehicle state may be used by a control component (e.g., a vehicle controller) for controlling the vehicle through a physical environment. For example, the control component may use these outputs of the DNN to determine a control profile (e.g., steering, decelerating, and/or accelerating) specific to the vehicle for controlling the vehicle through the physical environment.