Lane Position Control via Geometric Feedforward
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automated systems for controlling a vehicle's lateral lane position face conflicting requirements of high loop-gain for performance and low loop-gain for comfort, exacerbated by measurement delays and noise, leading to jerky steering and stability issues.
Innovation Solution
The method involves using a geometric model to generate a feedforward control signal based on inverse vehicle dynamics, combined with feedback control, to smoothly steer the vehicle to a desired lane position, reducing the need for high feedback loop-gain and minimizing control signal activity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If high loop-gain is used in the feedback controller, then controller performance is improved, but control signal activity increases causing jerky steering and reduced comfort
Solution Approach 1:
The patent applies preliminary action by calculating a feedforward steering signal that predicts the required steering input based on the geometric path model before the error occurs. This allows the controller to prepare the necessary control action in advance, reducing the need for high feedback gain and subsequent corrective actions that cause jerky steering movements.
Solution Approach 2:
The patent introduces a geometric path model as an intermediary between the desired lane position and the control signal. This model generates reference lateral position, velocity, and acceleration that serve as intermediate targets, allowing the feedback controller to operate with lower gain while still achieving accurate lane keeping through the coordinated feedforward-pathway.
2Speed
If high loop-gain is used in the feedback controller, then response speed is improved, but stability is reduced due to measurement delays and noise
Solution Approach 1:
By calculating the feedforward steering signal based on the geometric path model in advance, the system achieves fast response without relying on high feedback gain. The reference trajectory is computed proactively, allowing the controller to follow the desired path smoothly even with delayed or noisy measurements.
Solution Approach 2:
The patent uses feedback to correct deviations from the pre-calculated geometric path rather than to generate the entire control signal. This reduced-feedback approach with lower gain maintains stability by minimizing the amplification of measurement noise and delays, while still ensuring accurate tracking through the dominant feedforward path.
3Reliability
If derivative control is added to improve response, then performance is improved, but sensitivity to measurement noise and delay increases
Solution Approach 1:
The geometric path model calculates reference lateral acceleration and jerk proactively based on the desired path geometry, eliminating the need for derivative control that would differentiate noisy measurements. This preliminary calculation of reference dynamics provides smooth control signals without amplifying measurement noise.
Solution Approach 2:
The patent introduces the geometric path model as an intermediary that generates smooth reference trajectories with continuous derivatives. This intermediary layer filters out measurement noise before the control signal is generated, allowing the system to achieve responsive performance without directly differentiating noisy sensor data.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method and system for controlling the lateral lane position of a vehicle. A geometric model of a reference lateral path from an actual to a desired lateral lane position is provided. These steps are then performed for each control time point: determining a control error by comparing a reference lateral position, velocity and acceleration provided from the geometric model with an actual lateral position, velocity and acceleration; calculating a reference feedback steering angle based on the control error; calculating a total reference lateral acceleration by adding the reference lateral acceleration from the geometric model and a road-based reference lateral acceleration; calculating a reference feedforward steering angle by applying a model of the inverse vehicle dynamics on the total reference lateral acceleration; determining a steering angle error by comparing the actual steering angle with added reference feedforward and feedback steering angles, and providing a steering signal based on the steering angle error to a steering system (8).