Road Wetness Deep Learning Model for Autonomous Driving Decisions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately detecting road wetness, which can impact their operation, including sensor evaluation, wiper system engagement, and real-time and planned driving behavior.
Innovation Solution
A deep learning model is developed using on-board sensor signals and off-board environmental information to classify or regress road wetness, allowing autonomous vehicles to operate without relying on ground truth data during real-world driving.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ground truth data (water thickness measurements) are used to train the deep learning model, then measurement precision of road wetness is improved, but device complexity increases due to requiring specialized measurement equipment
Solution Approach 1:
The patent uses ground truth water thickness measurements during the training phase to create a deep learning model, but then replaces the need for physical measurement equipment during deployment by using the trained model to estimate road wetness from standard sensor data. The model copies the detection capability without requiring the physical measurement devices.
Solution Approach 2:
The patent replaces specialized mechanical measurement equipment (water thickness sensors) with a software-based deep learning model that processes data from standard vehicle sensors. This substitution eliminates the need for complex hardware while maintaining detection accuracy through algorithmic processing.
2Reliability
If multiple on-board sensors and off-board information sources are integrated, then reliability of road condition detection is improved, but device complexity increases
Solution Approach 1:
The patent merges data from multiple on-board sensors (cameras, lidars, radars) with off-board information (weather data, road graphs) into a unified deep learning model. This integration combines diverse data sources to improve reliability while the model architecture manages the complexity of processing multiple input streams.
Solution Approach 2:
The deep learning model serves multiple functions: it processes various sensor types (camera, lidar, radar), integrates off-board information, and outputs road wetness classification. This multi-functionality consolidates what would otherwise require separate systems into a single unified model.
3Speed
If deep learning model processing is performed in real-time, then responsiveness of autonomous vehicle operation is improved, but use of energy increases
Solution Approach 1:
The deep learning model is trained offline in advance using extensive ground truth data and sensor inputs. This preliminary training phase allows the model to be pre-optimized for real-time inference, reducing the computational energy required during actual vehicle operation while maintaining fast processing speeds.
Data Source
AI summary
The technology relates to using on-board sensor data, off-board information and a deep learning model to classify road wetness and/or to perform a regression analysis on road wetness based on a set of input information. Such information includes on-board and/or off-board signals obtained from one or more sources including on-board perception sensors, other on-board modules, external weather measurement, external weather services, etc. The ground truth includes measurements of water film thickness and/or ice coverage on road surfaces. The ground truth, on-board and off-board signals are used to build the model. The constructed model can be deployed in autonomous vehicles for classifying/regressing the road wetness with on-board and/or off-board signals as the input, without referring to the ground truth. The model can be applied in a variety of ways to enhance autonomous vehicle operation, for instance by altering current driving actions, modifying planned routes or trajectories, activating on-board cleaning systems, etc.