Action Reference Pipeline for Vehicle Trajectory Tree Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing planning systems in autonomous and semi-autonomous vehicles struggle to accurately predict future actions to avoid objects in the environment, leading to potential safety issues and inefficiencies in navigation.
Innovation Solution
Implementing a vehicle computing device with a model that generates multiple action trajectories using generators to modify baseline trajectories, combining them in various ways to create a tree structure for a tree search algorithm, optimizing computational resources and improving prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a tree search algorithm is used to determine vehicle trajectories, then the accuracy of action prediction is improved, but the computational time and complexity increase
Solution Approach 1:
The system pre-generates action references and trajectories before the tree search execution. Multiple trajectory generators create candidate trajectories in advance, which are then used as inputs for the tree search algorithm. This preliminary action reduces the computational burden during real-time decision making while maintaining prediction accuracy.
Solution Approach 2:
The trajectory generation process is divided into multiple independent generators, each responsible for creating specific types of trajectories (e.g., lane changes, turns, straight paths). These segmented generators operate in parallel to produce diverse action references that are then fed into the tree search, improving both efficiency and accuracy.
2Reliability
If multiple trajectory generators are combined to produce action trajectories, then the number of potential trajectories increases improving safety, but the device complexity increases
Solution Approach 1:
The trajectory generators are designed with universal interfaces and standardized output formats, allowing multiple generators to be combined without increasing proportional complexity. Each generator can produce multiple types of trajectories using the same underlying framework, reducing overall system complexity while maintaining high safety through diverse trajectory options.
Solution Approach 2:
Multiple trajectory generators are merged into a unified system that shares common resources such as environment models, vehicle dynamics models, and validation logic. This merging approach allows the system to produce a greater number of potential trajectories for improved safety while avoiding linear increases in complexity through shared infrastructure.
3Productivity
If parallel processing is implemented to optimize computational resources, then the processing speed is improved, but the device complexity increases
Solution Approach 1:
The system replaces complex sequential processing architecture with parallel processing that leverages standard multi-core CPU and GPU capabilities. By substituting the mechanical processing sequence with concurrent execution on available hardware threads, the system achieves higher processing speed while using conventional parallel computing infrastructure rather than specialized complex architectures.
Data Source
AI summary
Techniques for accurately predicting and avoiding collisions with objects detected in an environment of a vehicle are discussed herein. A vehicle computing device can implement a model to generate action references usable in a tree search to control the vehicle. For example, the model may represent one or more trajectory generators that modify a baseline trajectory in different ways to output a set of actions trajectories that are usable to define a tree structure. The set of action trajectories can be used by the vehicle computing device for predicting vehicle actions by the vehicle computing device to control the vehicle.