Autonomous Driving Path Planning Using A* Search and Polynomial Curves

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current path planning methods for autonomous driving require significant storage space, fail to meet vehicle nonholonomic constraints, and lack adaptability and scalability, particularly in complex traffic scenarios.

Innovation Solution

A path planning method that generates candidate sub-paths using environment perception and vehicle positioning information, performs collision detection, and employs the A* search algorithm to find optimal paths that meet vehicle constraints, thereby reducing storage requirements and improving adaptability and scalability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If Delaunay triangulation is used to generate virtual nodes for path planning, then path planning coverage is improved, but storage space requirement increases significantly

Engineering Contradiction:

Improvepath planning coverageVSAvoidstorage space

Core Design Contradiction:

Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent divides the path planning process into two independent modules: global path planning (road-level navigation) and local path planning (vehicle-level obstacle avoidance). This segmentation allows each module to use appropriate data structures - global planning uses simplified road networks while local planning uses polynomial curves - reducing overall storage requirements while maintaining comprehensive path planning coverage

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent replaces the need to store large numbers of virtual nodes from Delaunay triangulation with polynomial curve representations of paths. Instead of storing discrete virtual node coordinates, the system stores compact polynomial coefficients that can generate infinite path points, significantly reducing storage space while maintaining path planning flexibility

Inventive Principle:

Principle #26Copying

2Productivity

If graph searching with virtual nodes is used for path planning, then path finding capability is improved, but vehicle nonholonomic constraint satisfaction deteriorates

Engineering Contradiction:

Improvepath finding capabilityVSAvoidvehicle nonholonomic constraint satisfaction

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent changes the mathematical representation of paths from discrete virtual node connections to continuous polynomial curves (specifically quintic polynomials). This parameter change allows the path to be defined by smooth mathematical functions that can inherently satisfy vehicle kinematic constraints including nonholonomic constraints, while maintaining efficient path finding through coefficient optimization

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces the mechanical graph searching approach (discrete node-to-node traversal) with a mathematical optimization approach using polynomial fitting. This substitution eliminates the need to physically traverse virtual nodes and instead directly computes paths that satisfy vehicle constraints through mathematical optimization, improving both efficiency and constraint satisfaction

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Stability of the object's composition

If polynomial fitting is used for lane changing path planning, then path smoothness is improved, but scalability to different traffic scenarios deteriorates

Engineering Contradiction:

Improvepath smoothnessVSAvoidscalability to different traffic scenarios

Core Design Contradiction:

Stability of the object's compositionVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal path planning framework where polynomial curves serve multiple functions across different traffic scenarios. The same quintic polynomial formulation is used for lane changes, obstacle avoidance, intersections, and other scenarios, making the system scalable and adaptable while maintaining path smoothness through the inherent mathematical properties of polynomial functions

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS11460311B2Path planning method, system and device for autonomous driving

Publication Date: 2022.10.04 GUANGZHOU XIAOPENG MOTORS TECH CO LTD

AI summary

A path planning method, system, and device for autonomous driving are provided, where the method includes: acquiring environment perception information and vehicle positioning and navigation information, where the environment perception information includes obstacle information, roadside information, and lane line information, and the vehicle positioning and navigation information includes a vehicle pose and a target route; generating sub-paths according to the environment perception information and the vehicle positioning and navigation information, to obtain candidate sub-paths meeting vehicle constraints; performing a collision detection on the candidate sub-paths meeting the vehicle constraints, to obtain collision-free candidate sub-paths; performing a sub-path search on the collision-free candidate sub-paths by using an A* search algorithm; and obtaining, according to a result of the sub-path search, a local path of a vehicle.