Neural Network Digital Mapping for Occlusion-Aware 3D Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital map technologies face challenges in producing accurate three-dimensional images for autonomous vehicles, including occlusions, dynamic objects, and spatial deviations, which affect the precision required for autonomous navigation.
Innovation Solution
A neural network system is employed to process image data, applying weights to nodes to produce a digital map that accounts for occlusions, dynamic objects, and spatial deviations, and includes a Bayesian filter for predicting dynamic states, ensuring accurate representation of static and dynamic elements in the environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional digital map production methods are used, then the system complexity remains manageable, but the measurement precision and manufacturing precision required for autonomous vehicle navigation (within decimeter accuracy) cannot be achieved
Solution Approach 1:
The digital map is divided into multiple layers (base map layer, geometric map layer, semantic map layer, map priors layer, and real-time knowledge layer), each handling specific types of spatial information. This segmentation allows complex precision requirements to be addressed in specialized sub-systems rather than a monolithic structure.
Solution Approach 2:
A neural network is introduced as an intermediary processing system between raw sensor data and the final digital map output. The neural network accounts for occlusions, dynamic objects, and spatial deviations, transforming imperfect input data into high-precision map information without requiring complete redesign of the entire system.
2Volume of stationary object
If three dimensional images are produced from limited viewpoint data, then the volume of information represented increases, but occlusions and spatial deviations reduce the reliability of the representation
Solution Approach 1:
The neural network is pre-trained to recognize and compensate for common occlusion patterns and spatial deviations before map production occurs. By preparing the system in advance with learned models of these distortions, the reliability of three-dimensional representations is improved without requiring additional physical viewpoints.
Solution Approach 2:
The system incorporates feedback mechanisms where the neural network continuously refines its understanding of spatial relationships by comparing predicted positions with actual sensor data. This feedback loop allows the system to correct for occlusions and deviations iteratively, maintaining reliability despite limited viewpoint data.
3Adaptability or versatility
If dynamic objects are included in three dimensional images for the geometric map layer, then the real-time knowledge layer improves, but the manufacturing precision of static aspect representation deteriorates
Solution Approach 1:
The map system separates dynamic object information into a dedicated real-time knowledge layer, distinct from the geometric map layer that represents static aspects. This segmentation allows each layer to be optimized for its specific purpose without compromising the other.
Solution Approach 2:
The neural network acts as an intermediary that processes sensor data to distinguish between dynamic and static elements. It filters and categorizes information, directing static object data to the geometric map layer with high precision while routing dynamic object information to the real-time knowledge layer.
Data Source
AI summary
Information that identifies a location can be received. In response to a receipt of the information that identifies the location, a file can be retrieved. The file can be for the location. The file can include image data and a set of node data. The set of node data can include information that identifies nodes in a neural network, information that identifies inputs of the nodes, and values of weights to be applied to the inputs. In response to a retrieval of the file, the weights can be applied to the inputs of the nodes and the image data can be received for the neural network. In response to an application of the weights and a receipt of the image data, the neural network can be executed to produce a digital map for the location. The digital map for the location can be transmitted to an automotive navigation system.