Road Wetness Deep Learning Model for Autonomous Vehicle Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately classifying and responding to road wetness conditions due to limitations in existing sensor systems, which can impact driving operations and safety, especially when ground truth data is not readily available on-board.
Innovation Solution
A deep learning model is developed using on-board and off-board sensor data, including weather information and road graph data, to classify and regress road wetness without relying on ground truth sensors, enabling enhanced autonomous vehicle operation by altering driving actions, modifying routes, and activating cleaning systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If on-board sensors are used to detect road wetness, then real-time road condition information is obtained, but measurement precision is insufficient without ground truth sensors
Solution Approach 1:
The patent introduces an intermediary deep learning model that processes data from existing on-board sensors (lidar, radar, cameras) to infer road wetness conditions. This intermediary system bridges the gap between available sensor data and accurate wetness classification without requiring direct ground truth sensors on the vehicle, thereby maintaining measurement precision while avoiding additional hardware complexity
Solution Approach 2:
The patent creates a virtual copy of ground truth measurement capability through software-based deep learning models. Instead of physically installing ground truth sensors on each vehicle, the system replicates their measurement function by training models on labeled data and deploying them on standard on-board sensors, enabling accurate wetness detection through computational rather than physical means
2Measurement precision
If ground truth sensors are installed on-board, then measurement precision improves, but device complexity and cost increase
Solution Approach 1:
The patent performs preliminary action by collecting and labeling ground truth data during dedicated data collection phases, then using this pre-processed labeled data to train deep learning models. This preliminary preparation allows the models to achieve ground truth-level measurement precision when deployed on standard on-board sensors, eliminating the need to install actual ground truth sensors on production vehicles
Solution Approach 2:
The patent replaces expensive, complex ground truth sensors with inexpensive, widely available standard on-board sensors (lidar, radar, cameras). The cheap sensors are enhanced through software algorithms rather than hardware, achieving comparable measurement precision at fraction of the cost and complexity of dedicated ground truth sensing equipment
3Measurement precision
If multiple sensor types are integrated, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent merges data from multiple sensor types (lidar, radar, cameras) into a unified deep learning model that processes all inputs simultaneously. By combining sensor inputs at the model input layer rather than through complex intermediate fusion logic, the system achieves improved measurement precision through multi-sensor integration while keeping the overall system architecture relatively simple and manageable
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.