Iteratively Learning Controller for Vehicle Trajectory Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driving dynamics control systems for semi-automated or automated transportation vehicles rely on inaccurate models with fixed parameters, leading to inefficiencies in traversing planned trajectories due to fluctuating real-world conditions.
Innovation Solution
A method and device utilizing a control circuit with two degrees of freedom, comprising a pilot controller, a controller, and an iteratively learning controller, which classifies planned trajectories, adjusts manipulated variable profiles based on control errors and recorded variables, and optimizes control behavior through machine learning and database updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a control circuit with fixed model and fixed parameter set is used, then the device complexity is reduced, but the reliability deteriorates due to inaccurate models not matching real current parameters
Solution Approach 1:
The patent applies dynamics by transforming the fixed control model into an adaptive system that continuously learns and adjusts its parameters. The iteratively learning controller modifies the manipulated variable profile based on recorded control errors, enabling the system to adapt to changing real-world conditions while maintaining reasonable structural complexity.
Solution Approach 2:
The patent implements feedback by recording control errors during trajectory traversal and using these errors to adjust the manipulated variable profile in subsequent operations. This closed-loop learning mechanism allows the system to improve its accuracy over time without requiring a completely complex restructured control circuit.
2Manufacturing precision
If model parameters are estimated using observers or filters, then the manufacturing precision is improved, but the loss of time increases due to continuous parameter estimation processes
Solution Approach 1:
The patent applies preliminary action by pre-storing manipulated variable profiles for different trajectory classes in a database. When a trajectory is classified, the corresponding pre-computed profile is retrieved and adjusted based on recorded errors, avoiding the need for continuous real-time parameter estimation and reducing computation time while maintaining precision.
Solution Approach 2:
The patent segments the control approach by dividing trajectories into different classes and storing specific manipulated variable profiles for each class. This segmentation allows the system to select appropriate pre-computed profiles based on trajectory type, reducing the need for continuous parameter estimation across all scenarios.
3Device complexity
If the same manipulated variable profile is used for all trajectories, then the device complexity is reduced, but the adaptability deteriorates due to fluctuating real-world conditions
Solution Approach 1:
The patent applies dynamics by implementing an iteratively learning controller that adapts the manipulated variable profile based on the specific trajectory class and recorded control errors. The system dynamically adjusts the profile for each trajectory type while maintaining a structured database approach, achieving adaptability without excessive complexity.
Solution Approach 2:
The patent applies local quality by storing and applying different manipulated variable profiles for different trajectory classes. Each trajectory class receives a customized profile tailored to its specific characteristics, allowing local optimization for each trajectory type while maintaining an organized database structure.
Data Source
AI summary
A method for driving dynamics control for a transportation vehicle, wherein a manipulated variable of the driving dynamics is controlled by a control circuit having two degrees of freedom, consisting of a pilot control and a controller, to drive through a planned trajectory, wherein the control circuit has an iteratively learning controller which cyclically repeats classifying the planned trajectory by a classification device, retrieving a manipulated variable profile for the iteratively learning controller from a database based on the classification, recording a control fault of the control circuit and/or a manipulated variable of the controller when driving through the planned trajectory by a memory, and adapting the manipulated variable profile of the iteratively learning controller based on the recorded control fault and/or the recorded manipulated variable of the controller. Also disclosed is an associated device.