Autonomous Vehicle Trajectory Planning With Gradient-Based Collision Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately predicting collisions with dynamic obstacles and experience computational delays when processing polynomial expressions for path planning, leading to potential safety issues in real-time control operations.
Innovation Solution
A system and method for trajectory planning of autonomous vehicles that incorporates a human driving behavior model, predicts the trajectory of moving objects, and uses a gradient descent algorithm to evaluate and optimize path options, reducing computational load and enhancing safety by providing multiple path options and ranking them based on scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If polynomial expressions or mathematical functions are used to represent spatial information of the driving path, then the autonomous vehicle can perform control operations to avoid stationary obstacles, but the computational load increases causing unsafe delays in real-time control operations
Solution Approach 1:
The patent transforms the path planning problem from polynomial fitting to gradient-based optimization by changing the mathematical parameters from coefficients of polynomial expressions to gradient vectors (dx/ds, dy/ds). This parameter transformation enables real-time computation while maintaining collision avoidance capability, as gradient calculations are computationally simpler than polynomial fitting and optimization.
2Reliability
If multiple path options are generated and evaluated using a ranking system, then the best trajectory can be found, but the computational complexity increases
Solution Approach 1:
The patent segments the path planning process into distinct gradient-based optimization steps rather than evaluating multiple complete polynomial path options. By dividing the problem into incremental gradient descent steps that optimize the trajectory continuously, the system achieves high-quality trajectory optimization with reduced computational complexity compared to generating and ranking multiple complete paths.
Data Source
AI summary
A system and method for path planning of autonomous vehicles based on gradient are disclosed. A particular embodiment includes: executing a path planning module by use of a data processor; receiving, from a remote data source by use of the data processor, ancillary data to modify operation of the path planning module based on a context in which an autonomous vehicle is operating; generating, by use of the data processor, a trajectory comprising a plurality of waypoints; and causing the autonomous vehicle to follow a path conforming to the plurality of waypoints.