Autonomous Driving Reference Line Smoothing via QP Spline and Spiral Curvature Optimization
Solution Overview
Problem
Conventional trajectory planning techniques for autonomous driving vehicles rely on raw reference lines generated from map data, which often lack smoothness, leading to unstable and oscillating trajectories due to high curvatures, particularly in turns.
Innovation Solution
A computer-implemented method using a Quadratic Programming (QP) spline smoothing method followed by a spiral smoothing method to identify and optimize segments with excessive curvature, generating a smoothed reference line that satisfies specific constraints, thereby improving trajectory planning for autonomous driving vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If a raw reference line is directly generated from map data points, then the trajectory planning process is simple and fast, but the reference line lacks smoothness and causes unstable and oscillating trajectories
Solution Approach 1:
The reference line is divided into multiple segments, with each segment represented by a polynomial function. This segmentation allows independent optimization of each segment while maintaining overall continuity, resolving the contradiction between smoothing quality and computational complexity.
Solution Approach 2:
The patent applies curvature constraints to ensure smooth transitions between segments. By controlling the curvature continuity (G2 continuity) at segment junctions, the reference line achieves enhanced smoothness without requiring excessive computational resources.
2Stability of the object's composition
If conventional smoothing methods are applied to the reference line, then some smoothness is improved, but high curvature segments remain that cause oscillating trajectories
Solution Approach 1:
The patent applies different smoothing strategies to different segments of the reference line. High-curvature segments are identified and processed with specialized polynomial fitting that prioritizes curvature continuity, while low-curvature segments use standard smoothing. This local differentiation resolves the contradiction between overall smoothness and local precision.
Solution Approach 2:
The patent adjusts polynomial degree and curvature constraints as parameters to achieve optimal smoothing. By dynamically selecting polynomial orders (e.g., cubic, quartic, quintic) and adjusting curvature weights in the optimization objective, the system achieves precise curvature continuity without over-smoothing critical segments.
3Stability of the object's composition
If multiple smoothing iterations are performed to achieve high curvature continuity, then trajectory stability improves, but computational time and processing resources increase
Solution Approach 1:
The patent performs preliminary identification of high-curvature segments before applying intensive smoothing. By pre-processing the reference line to locate critical segments, the system avoids unnecessary computation on already-smooth portions, significantly reducing total processing time while maintaining trajectory stability.
Solution Approach 2:
The patent applies full smoothing effort only to segments where it is most needed (high-curvature segments), rather than uniformly processing the entire reference line. This partial action approach achieves the necessary curvature continuity with minimal computational overhead.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
In one embodiment, an exemplary method includes the operations of receiving a raw reference line representing a route from a first location to a second location associated with an autonomous driving vehicle (ADV); and smoothing the raw reference line using a Quadratic programming (QP) spline smoother to generate a smoothed reference line. The method further includes the operations of identifying one or more segments on the smoothed reference line, each of the identified reference line segments including a curvature that exceeds a predetermined size; and smoothing each of the one or more identified reference line segments using a spiral smoother, including optimizing each identified curvature in view of a set of constraints, such that an output of the objective function reaches a minimum value while the set of constraints are satisfied; and controlling the ADV using the smoothed reference line.