Velocity Trajectory Generation for Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing velocity planning methods for driverless vehicles often result in velocity trajectories that do not match actual driving scenes, leading to data redundancy and increased calculation complexity in subsequent optimization algorithms.
Innovation Solution
A method and apparatus for generating velocity trajectories by setting starting, intermediate, and terminal point times, and using polynomial functions to generate first and second velocity sub-trajectories based on initial and terminal velocities, with continuous acceleration and jerk changes, to align with human driving characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If velocity-time space is divided into grids and multiple trajectories are planned in each grid, then the number of velocity trajectories increases, but data redundancy and calculation complexity increase
Solution Approach 1:
The velocity trajectory is segmented into multiple sub-trajectories based on different road section types (e.g., acceleration sections, deceleration sections, constant speed sections). Each sub-trajectory is generated using appropriate polynomial functions tailored to its specific driving scenario, avoiding the need to generate all possible trajectories in a grid-based approach.
Solution Approach 2:
Different polynomial function types are applied to different sections of the velocity trajectory based on local driving requirements. For example, third-order polynomials are used for acceleration sections, while fifth-order polynomials are used for deceleration sections where smoother transitions are needed. This localized approach generates fewer, more relevant trajectories.
2Adaptability or versatility
If multiple velocity trajectories are generated to cover various driving scenarios, then adaptability improves, but the amount of data to be processed increases
Solution Approach 1:
The patent changes the parameters of polynomial functions (order, coefficients) based on the specific driving scenario and road section characteristics. By adjusting polynomial order and coefficients according to acceleration thresholds, time intervals, and velocity constraints, the system generates trajectories that are adapted to specific driving needs rather than generating all possible trajectories.
Solution Approach 2:
The velocity trajectory generation is made dynamic by continuously adjusting polynomial parameters based on real-time driving conditions, road section types, and velocity constraints. This dynamic approach ensures that only relevant trajectories are generated for each specific driving scenario, reducing data redundancy while maintaining adaptability.
3Stability of the object's composition
If polynomial functions are used to generate velocity sub-trajectories, then smooth continuous acceleration and jerk are achieved, but computational complexity increases
Solution Approach 1:
The velocity trajectory is divided into multiple sub-trajectories, each governed by a specific polynomial function. This segmentation allows the use of simpler polynomial orders in sections where high smoothness is not critical, while applying higher-order polynomials only where needed (e.g., deceleration sections), thereby reducing overall computational complexity while maintaining acceleration continuity.
Data Source
AI summary
A velocity trajectory generation method and apparatus, and a storage medium are provided. The method includes: setting starting point time, intermediate point time and terminal point time of a velocity trajectory to be generated, wherein the velocity trajectory includes a first velocity sub-trajectory and a second velocity sub-trajectory; generating multiple first velocity sub-trajectories corresponding to a starting point velocity between the starting point time and the intermediate point time, according to the starting point velocity and multiple intermediate point velocities; and generating, for each of the first velocity sub-trajectories, multiple second velocity sub-trajectories between the intermediate point time and the terminal point time according to multiple terminal point velocities. A velocity trajectory which is more in line with a human driving characteristic may be generated, the comfort level of a passenger may be improved, and an algorithm may be simplified.