Vehicle Motion Planning With Double-Tree Path Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current path planning methods for autonomous vehicles, such as RRT-based methods, are computationally expensive and inefficient, especially in complex environments like parking garages, leading to impractical solutions and paths that may not align with human intuition due to probabilistic sampling, resulting in high computational costs and memory requirements.
Innovation Solution
A deterministic motion planning method using a doubletree graph construction approach, where an initial tree and a target tree are constructed to explore the state space efficiently, with nodes expanded based on cost and similarity of primitive motions, reducing unnecessary sampling and improving computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If probabilistic sampling methods (RRT) are used for path planning, then the method can handle high-dimensional configuration spaces, but the computational cost and memory requirements become excessively high
Solution Approach 1:
The configuration space is divided into two separate trees: a forward tree growing from the initial state and a backward tree growing from the target state. Each tree independently explores the configuration space, reducing the computational burden compared to exploring the entire high-dimensional space with a single probabilistic method. The path is constructed by connecting nodes from both trees, effectively segmenting the complex path planning problem into two more manageable subproblems.
2Reliability
If probabilistic sampling methods are used for path planning, then the method can explore the state space, but the paths generated do not align with human intuition due to random sampling nature
Solution Approach 1:
Instead of randomly sampling the configuration space as in traditional RRT methods, the patent uses deterministic graph-based expansion from both the initial and target states. Nodes are added based on geometric relationships and cost functions rather than random sampling, which produces paths that follow more predictable, human-like patterns while still thoroughly exploring the state space through systematic graph traversal.
3Manufacturing precision
If graph-based methods (A*, D*) are used for path planning, then optimality can be guaranteed under certain circumstances, but the complexity grows exponentially with configuration space dimension
Solution Approach 1:
The patent transforms the high-dimensional path planning problem into a two-dimensional graph search problem by representing the configuration space as a graph where nodes are configurations and edges are valid transitions. The dual-tree approach further reduces complexity by working in two opposite directions simultaneously, effectively adding a temporal dimension to the search process and avoiding exponential growth with configuration space dimension.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
A system for controlling a movement of a vehicle from an initial state of the vehicle and a target state of the vehicle constructs a graph having multiple nodes defining states of the vehicle and including an initial node defining the initial state of the vehicle and a target node defining the target state of the vehicle and determines a path through the graph connecting the initial node with the target node. The system determines the graph using doubletree construction with an initial tree of nodes originating at the initial node and a target tree of nodes originating at the target node. The doubletree construction is configured to select an expandable node in the initial tree or the target tree based on a cost of the expandable node, and expand the graph by adding a child node connected to the expandable node with an edge defined by a collision free primitive motion, such that a cost of the child node is less than the cost of the expandable node.