Dynamic Trajectory Prediction for Autonomous Vehicle Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems lack the ability to accurately predict the driving intentions of moving vehicles, which is crucial for autonomous decision-making and infrastructure planning, as they rely on stationary object detection methods that fail to discern driver intent.
Innovation Solution
A method that generates projected vehicle information by receiving remote vehicle spatial state information, identifying vehicle transportation network data, and updating likelihoods of lane following based on deviations, allowing a host vehicle to traverse the network using predicted trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stationary object detection methods are used, then the system is simple to implement, but the ability to predict driving intentions is lost
Solution Approach 1:
The system transitions from static object detection to dynamic trajectory prediction by continuously updating probability values based on sequential spatial state information and deviations from predicted paths, enabling the system to adapt to changing vehicle intentions
Solution Approach 2:
The system employs feedback mechanisms by comparing actual vehicle positions against predicted trajectories and updating probability values accordingly, allowing continuous refinement of driving intention predictions based on observed deviations
2Measurement precision
If remote vehicle spatial state information is continuously monitored and processed, then driving intention prediction accuracy improves, but computational resources and processing time increase
Solution Approach 1:
The system performs preliminary trajectory predictions using initial probability values before actual vehicle positions are fully observed, allowing proactive preparation of prediction models and reducing real-time computational burden
Solution Approach 2:
The system focuses computational resources on predicting only the relevant portion of vehicle trajectories needed for immediate navigation decisions, rather than computing complete long-term paths, optimizing the balance between accuracy and processing efficiency
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A host vehicle receives remote vehicle spatial state information for a remote vehicle and identifies vehicle transportation network information representing a portion of a transportation network based on that information. At least one initial probability value is generated based on comparing the spatial state information and the transportation network information at an initial time point, each initial probability value indicating a likelihood that the remote vehicle is following a lane within the transportation network. A deviation between adjacent values for the spatial state information relative to the transportation network information is generated for a plurality of time points. For each single lane and deviation, the likelihood that the remote vehicle is following the lane using a new probability value based on the deviation is updated, a trajectory using the updated likelihood is generated, and the host vehicle traverses the transportation network using the transportation network information and trajectory.