Variable Boundary Estimation for Autonomous Vehicle Path Planning

Solution Overview

Problem

Current path planning systems for autonomous vehicles face challenges in efficiently generating paths that balance passenger comfort, safety, and collision avoidance while maintaining proximity to the lane centerline and avoiding obstacles with a buffer, especially in dynamic environments.

Innovation Solution

The system employs a method for variable boundary estimation using a vehicle kinematic model, incorporating vehicle parameters and sensor data to determine optimal paths that account for drivability, comfort, and obstacle avoidance, by converting vehicle configurations from Cartesian to Frenet frames and using a perception and planning system to generate and optimize paths in the SL-coordinate system.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the vehicle maintains a buffer distance from obstacles to ensure safety, then collision avoidance is improved, but the path may deviate from the lane centerline

Engineering Contradiction:

Improvecollision avoidanceVSAvoidlane centerline proximity

Core Design Contradiction:

ReliabilityVSShape

Solution Approach 1:

The cost function applies different weighting factors to different spatial regions: higher weight for lane centerline adherence in open areas, and higher weight for obstacle buffer maintenance when obstacles are detected. This local adaptation resolves the contradiction by prioritizing different objectives in different contexts.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The path planning system dynamically adjusts the cost function parameters based on real-time environmental conditions, obstacle positions, and vehicle state. The weighting between lane centerline proximity and obstacle buffer distance is not fixed but adapts dynamically, allowing the system to resolve the contradiction based on current driving context.

Inventive Principle:

Principle #15Dynamics

2Reliability

If the path planning considers multiple factors including comfort and safety, then path quality is improved, but the computational complexity increases

Engineering Contradiction:

Improvepath safety and comfortVSAvoidpath planning system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system uses a continuous cost function with adjustable parameters (weighting factors) that can be tuned to balance different objectives. By changing parameters rather than implementing complex discrete decision logic, the system achieves high-quality path planning with manageable computational complexity.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces complex mechanical decision-making systems with a mathematical cost function evaluation approach. Instead of using multiple separate algorithms for different path constraints, a unified cost function evaluates all factors simultaneously, reducing overall system complexity while maintaining comprehensive path quality assessment.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Ease of operation

If the vehicle follows the lane centerline closely for smooth paths, then passenger comfort is improved, but the ability to maintain buffer from obstacles is reduced

Engineering Contradiction:

Improvepassenger comfortVSAvoidobstacle buffer maintenance

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The cost function applies different weighting factors to different spatial regions: higher weight for lane centerline adherence in open areas, and higher weight for obstacle buffer maintenance when obstacles are detected. This local adaptation resolves the contradiction by prioritizing different objectives in different contexts.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The path planning system dynamically adjusts the cost function parameters based on real-time environmental conditions, obstacle positions, and vehicle state. The weighting between lane centerline proximity and obstacle buffer distance is not fixed but adapts dynamically, allowing the system to resolve the contradiction based on current driving context.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS11273837B2Variable boundary estimation for path planning for autonomous driving vehicles

Publication Date: 2022.03.15 BAIDU USA LLC

AI summary

In one embodiment, a method of generating a path for an autonomous driving vehicle (ADV) is disclosed. The method includes obtaining vehicle configuration of the ADV. The vehicle configuration includes a longitudinal state of the ADV and a lateral state of the ADV relative to a discretized point along a reference line. The method further includes estimating one or more boundaries for the lateral state of the ADV with respect to the longitudinal state of the ADV. The estimation of the boundaries includes obtaining vehicle parameters and sensor data of the ADV, using a vehicle kinematic model to estimate the boundaries based on the vehicle parameters and the sensor data, and outputting the estimated boundaries. The method further includes generating an optimal path to control the ADV based on the vehicle configuration and the estimated boundaries.