Self-Supervised Vehicle Control Models With Graph Scene Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous vehicles require large amounts of labeled data for training machine-learning based models, which can be resource intensive and difficult to update based on changes to models.
Innovation Solution
A self-supervised machine learned model is developed to generate a representation of an environment, allowing for reduced redundancy in computing resources and improved performance by using a backbone model that can be trained in a self-supervised manner and adapted for specific tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If large amounts of labeled data are used for training machine-learning models, then model performance is improved, but resource consumption and training difficulty increase
Solution Approach 1:
The system employs self-supervised learning where the model learns to represent environmental data by predicting missing or masked portions of the input data itself, without requiring external labeled data. The model serves its own training needs by generating its own supervision signals from raw sensor data, eliminating the need for extensive manual labeling resources
Solution Approach 2:
A backbone model is pre-trained in a self-supervised manner on large volumes of raw environmental data before being adapted for specific downstream tasks. This preliminary training phase establishes a robust general representation that can be fine-tuned for specific applications, reducing the resources needed for task-specific training
2Adaptability or versatility
If multiple separate models are used for different tasks, then task specialization is improved, but computing resource redundancy increases
Solution Approach 1:
The backbone model is designed as a universal representation learner that can be adapted to multiple different downstream tasks through fine-tuning or feature extraction. This single multi-functional model replaces what would otherwise require multiple separate task-specific models, reducing computing resource redundancy while maintaining task specialization capabilities
Solution Approach 2:
The system separates the learning process into two distinct phases: (1) a backbone model trained in a general self-supervised manner to learn environmental representations, and (2) task-specific adaptation phases where the pre-trained backbone is fine-tuned for specific downstream tasks. This segmentation allows resource-efficient general learning followed by targeted task adaptation
Data Source
AI summary
There is provided a system configured to receive data associated with a vehicle operating within an environment; generate, based at least in part on the data, a graph comprising a plurality of nodes, a node of the plurality associated with one or more of a vehicle operating in the environment, a road feature, an additional vehicle, or a pedestrian; input the graph into a self-supervised machine learned model comprising an encoder, wherein the machine learned model is trained to output a representation associated with the node; receive, from the self-supervised machine learned model, a representation associated with the node; and transmit the representation to a downstream machine learned model trained to output control data based at least in part on the representation, wherein the control data is configured to control the vehicle or another vehicle.