Safe Driving Path Learning for Autonomous Adversity Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face challenges in navigating through adversity conditions such as road work zones, obscured or missing road markings, and obstacles, as they are often designed to make logic-based and cost-based decisions, which may not account for all scenarios, leading to potential errors.
Innovation Solution
Training machine learning models using vehicle sensor data, including image and inertial measurement data, to predict and prescribe safe navigation trajectories, detect adversity conditions, and provide control instructions for autonomous vehicles to navigate through challenging situations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If logic-based and cost-based decisions are used in autonomous driving systems, then decision-making speed is improved, but reliability in adversity conditions deteriorates
Solution Approach 1:
The patent introduces machine learning models as an intermediary between sensor data and decision-making. These models are trained on extensive driving data including adversity conditions, enabling the system to learn complex patterns and make reliable decisions in tricky situations without sacrificing decision-making speed. The ML models act as a mediator that translates raw sensor data into informed navigation decisions.
2Reliability
If machine learning models are trained on extensive driving data to improve reliability, then reliability in adversity conditions is improved, but device complexity increases
Solution Approach 1:
The patent applies preliminary action by training machine learning models offline on extensive driving data before deployment. During actual autonomous driving operation, the pre-trained models make predictions without requiring complex real-time computations. This shifts the computational complexity from the operating phase to the training phase, maintaining reliability while reducing operational device complexity.
3Adaptability or versatility
If autonomous driving systems navigate through all possible scenarios, then coverage of driving conditions is improved, but loss of time in data processing increases
Solution Approach 1:
The patent uses parameter changes by training machine learning models on diverse driving data representing various conditions, scenarios, and environments. The models learn to generalize from this training data, enabling them to adapt to new situations without requiring explicit programming for each scenario. This reduces real-time processing time while maintaining broad coverage of driving conditions.
Data Source
AI summary
Systems and methos for learning safe driving paths in autonomous driving adversity conditions can include acquiring, by a computer system, vehicle sensor data for a plurality of driving events associated with one or more autonomous driving adversity conditions. The vehicle sensor data can include, for each driving event, corresponding image data depicting surroundings of the vehicle and corresponding inertial measurement data. The computer system can acquire, for each driving event of the plurality of driving events, corresponding vehicle positioning data indicative of a corresponding trajectory followed by the vehicle, and train, using the vehicle sensor data and the vehicle positioning data, a machine learning model to predict navigation trajectories during the autonomous driving adversity conditions.