Autonomous Vehicle Trajectory Prediction via Contextual RNN Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in predicting the trajectories of nearby vehicles due to complex driving environments, which can lead to difficulties in making safe and appropriate driving decisions.
Innovation Solution
A system that emulates human thought processes by perceiving the driving environment, using machine learning models like recurrent neural networks to predict trajectories based on vehicle-independent information such as time, driving conditions, and traffic information, allowing the autonomous vehicle to make informed decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If autonomous vehicles use onboard sensors to navigate and make driving decisions, then the vehicle can operate with minimal human interaction, but it becomes difficult to determine or identify the next maneuver the vehicle will make in complex environments
Solution Approach 1:
The system performs preliminary actions by predicting multiple possible trajectories for surrounding vehicles before the autonomous vehicle makes its driving decision. The trajectory prediction module pre-calculates likely paths of other vehicles based on their current states and environmental context, providing advance information that helps the autonomous vehicle plan safe maneuvers without requiring real-time complex reasoning about other drivers' intentions
Solution Approach 2:
The patent introduces trajectory prediction as an intermediary layer between sensor data collection and driving decision-making. Instead of directly processing raw sensor data to make decisions, the system uses trajectory prediction models (including recurrent neural networks) as intermediaries that translate observed vehicle states into predicted future paths, which then inform the autonomous vehicle's maneuver selection
2Measurement precision
If the autonomous vehicle uses machine learning models to predict trajectories based on vehicle-independent information, then the prediction accuracy improves, but the computational complexity and system complexity increase
Solution Approach 1:
The system extracts and focuses only on vehicle-independent information (such as road geometry, traffic signs, environmental context) that is relevant for trajectory prediction, separating these key features from the complete set of available sensor data. This extraction approach improves prediction accuracy by concentrating on discriminative features while reducing the dimensionality of input data that the machine learning models must process
Solution Approach 2:
The patent employs recurrent neural networks that learn from training data to create computational models copying human-like trajectory prediction capabilities. The RNNs are trained on extensive datasets of driving scenarios and replicate human drivers' ability to predict vehicle paths, enabling the autonomous vehicle to make accurate predictions without requiring explicit programming of complex driving logic
Data Source
AI summary
A surrounding environment of an autonomous vehicle is perceived to identify one or more vehicles nearby. For each of the identified vehicles, based on a current location of the identified vehicle, vehicle-independent information is obtained to determine context surrounding the identified vehicle, where the vehicle-independent information includes vehicle surrounding information that defines physical constraints imposed on the identified vehicle. For each of the identified vehicles, one or more trajectories for the identified vehicle are predicted based at least in part on the vehicle-independent information associated with the identified vehicle. The autonomous vehicle is controlled based on the one or more predicted trajectories of the one or more identified vehicles.