Heavy-Duty Vehicle Trajectory Planning With Deadlock Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Planning guided trajectories for heavy-duty vehicles is challenging due to the risk of traffic deadlocks, especially in areas with traffic constraints, as these vehicles require complex operation and specialized skills to extricate from deadlock situations, and existing technologies fail to effectively predict and prevent deadlocks in autonomous or semi-autonomous operations.
Innovation Solution
A traffic planner with a deadlock prediction pre-processor that uses a trained classifier model to predict potential deadlocks by analyzing vehicle trajectories and occupancy states, allowing for dynamic adjustment of trajectories to avoid congestion and optimize vehicle movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple heavy-duty vehicles follow guided trajectories through traffic constrained locations, then vehicle fleet productivity is improved, but traffic deadlock risk increases
Solution Approach 1:
The deadlock prediction pre-processor performs preliminary analysis of traffic situations before deadlocks occur. It classifies current traffic states and predicts future deadlock risks, allowing the traffic planner to take preventive actions by adjusting vehicle trajectories in advance to avoid predicted deadlock scenarios.
Solution Approach 2:
The system implements continuous feedback by repeatedly classifying traffic situations as vehicles move through the site. The deadlock classifier continuously monitors vehicle positions and trajectory intersections, providing real-time predictions that feed back to the traffic planner for dynamic trajectory adjustments.
2Reliability
If trajectory planning is optimized to prevent deadlocks, then traffic reliability is improved, but computational complexity increases
Solution Approach 1:
The traffic situation is segmented into discrete classifiable states that the deadlock classifier can process. By dividing the continuous traffic environment into distinct scenarios (deadlock-prone vs. safe), the system simplifies the computational task while maintaining effective deadlock prediction capability.
Solution Approach 2:
The deadlock classifier is pre-trained on traffic situation data to recognize deadlock patterns. This preliminary training phase separates the complex computational work from real-time operation, allowing fast classification during actual traffic planning without requiring complex computations during critical decision moments.
3Loss of information
If real-time deadlock prediction is implemented, then traffic situation awareness is improved, but processing time increases
Solution Approach 1:
The deadlock classifier is pre-trained offline on comprehensive traffic situation data, preparing it to rapidly classify real-time scenarios. This preliminary training transfers computational burden from real-time operation to an initial setup phase, enabling fast processing during actual traffic management.
Solution Approach 2:
The system uses a simplified classification model that captures essential deadlock patterns without requiring full complex simulations. This copied approach to traffic analysis provides sufficient awareness for planning purposes while maintaining computationally efficient processing speeds.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Various methods and related aspects are disclosed to determine if a traffic situation comprising a plurality of heavy-duty vehicles following a plurality of guided vehicle trajectories in a defined area will lead to a vehicle deadlock. For example, one method comprises generating a traffic situation representation for the defined area comprising, for each of the plurality of vehicles, an initial vehicle trajectory segment location and a subsequent deadlock status, inputting the traffic situation representation to a traffic situation deadlock classifier which has been trained to output a traffic situation deadlock classification based on an input representation of a traffic situation in the defined area, and, based on the input traffic situation representation, generating a predicted deadlock classification of a subsequent traffic situation for that input traffic situation representation.