Autonomous Vehicle Lateral Control to Prevent Trajectory Overshoot
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional lateral controllers for autonomous vehicles are not well adapted to navigate complex and dynamic driving scenarios, particularly for larger vehicles like trucks, leading to potential overshooting of target trajectories.
Innovation Solution
The implementation of a 'catch and release' controller paradigm that includes a 'catch' phase for quickly aligning the vehicle with the target trajectory and a 'release' phase to maintain alignment without crossing it, using a surrogate trajectory, curvature targets, and PID functions to generate steering commands based on lateral and heading errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional lateral controllers use predefined rules and models, then the control system is simple to implement, but the controller cannot adapt well to complex and dynamic driving scenarios
Solution Approach 1:
The lateral controller transitions from static predefined rules to dynamic adaptive control by implementing two distinct operational phases (catch phase and release phase) that automatically switch based on real-time trajectory alignment conditions. The controller dynamically adjusts its behavior by calculating different curvature targets for each phase, enabling adaptation to varying driving scenarios without increasing structural complexity
Solution Approach 2:
The controller changes the curvature target parameter based on the operational phase. During the catch phase, the curvature target is designed to quickly reduce lateral error, while during the release phase, the curvature target is adjusted to prevent overshooting. This parameter transformation allows the same controller structure to handle different driving conditions effectively
2Speed
If the vehicle steers aggressively towards the target trajectory to reduce lateral error quickly, then the response speed is improved, but the vehicle may overshoot the target trajectory
Solution Approach 1:
The trajectory tracking process is segmented into two distinct phases: the catch phase for rapid error reduction and the release phase for precise alignment. This segmentation allows the system to optimize for speed during the catch phase and for reliability during the release phase, eliminating the trade-off between these two objectives
Solution Approach 2:
The controller implements periodic switching between catch and release phases based on trajectory alignment conditions. The system transitions from aggressive steering (catch phase) to conservative alignment (release phase) in a rhythmic manner, ensuring both rapid response and reliable trajectory following without overshooting
3Productivity
If the lateral controller uses a single control mode, then the control logic is simple, but the controller cannot handle both rapid alignment and precise trajectory following
Solution Approach 1:
The control logic dynamically adapts by implementing two distinct control modes (catch mode and release mode) that activate based on real-time trajectory alignment conditions. The system automatically transitions between modes to optimize performance, achieving both rapid alignment and precise trajectory following without requiring complex multi-mode control architecture
Data Source
AI summary
Systems and methods of automatic correction of map data for autonomous vehicle navigation are disclosed. One or more processors of an autonomous vehicle can determine a surrogate trajectory for navigating the autonomous vehicle towards a target trajectory; generate a curvature target for the autonomous vehicle based on the surrogate trajectory, a lateral error, and a velocity of the autonomous vehicle, the curvature target defined in part by at least three points; and navigate the autonomous vehicle according to the curvature target.