Variable Time-Step Tree Search for Vehicle Trajectory Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learned models for predicting actions in robotic devices, such as autonomous vehicles, face inefficiencies when using fixed step sizes for searching decision trees, leading to suboptimal processing resource allocation and reduced accuracy in predicting interactions with objects.
Innovation Solution
Implementing a machine learned model that outputs variable step sizes for searching a decision tree encoded with vehicle, object, and environment information, allowing for more efficient allocation of processing resources by dedicating more resources to earlier steps and less to subsequent longer steps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fixed step sizes are used for searching decision trees, then processing resource allocation is simplified, but prediction accuracy is reduced due to suboptimal resource allocation
Solution Approach 1:
The patent implements dynamic step size adjustment in the tree search algorithm, where the step size changes based on the current search state and remaining computational budget. This allows the system to allocate more resources to critical early decisions and fewer resources to later decisions, improving prediction accuracy while adapting resource allocation to actual needs rather than using fixed allocations throughout the search process.
2Measurement precision
If more processing resources are dedicated to earlier steps of tree search, then prediction accuracy improves, but computational complexity increases
Solution Approach 1:
The patent changes the parameter of step size from a fixed value to a variable that is dynamically adjusted during the tree search process. By modifying this parameter based on the search state and computational budget, the system achieves better prediction accuracy through improved resource allocation without permanently increasing computational complexity, as the complexity adapts to the problem instance rather than being uniformly high across all cases.
3Productivity
If variable step sizes are used for tree search, then computational resource efficiency is improved, but implementation complexity increases
Solution Approach 1:
The patent introduces dynamic step size adjustment that adapts during the tree search process based on the current state and remaining computational budget. This dynamic approach improves computational resource efficiency by allocating resources more effectively across different stages of the search, while the adaptation logic is designed to manage implementation complexity through systematic decision rules rather than arbitrary complexity.
Data Source
AI summary
Techniques for implementing a machine learned model to determine variable step sizes for searching a decision tree are described herein. For example, the techniques may include a computing device implementing a tree search algorithm to search the decision tree comprising nodes representing different possible object actions and/or vehicle actions at a future time. A search of the decision tree can be performed using multiple steps associated with different time periods to output a vehicle trajectory for the vehicle. The vehicle trajectory can be sent to a vehicle computing device for consideration during vehicle planning, which may include simulation.