Vehicle Path Tracking Observer Using Speed-Adaptive Gain Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing state observer devices struggle to accurately estimate the trajectory tracking state vector for autonomous vehicles, particularly at varying speeds, as they are often limited to simple systems and fail to adapt to complex dynamics.
Innovation Solution
An observer device that generates an estimated trajectory tracking state vector in real time by adjusting the temporal variation of the estimated state vector using a weighted summation of estimation gain matrices optimized for different speed ranges, with weighting coefficients that change based on the current speed, ensuring accurate tracking across a wide range of speeds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a simple observer structure is used, then the device complexity is reduced, but the measurement precision of the state vector estimation deteriorates
Solution Approach 1:
The observer structure is segmented into multiple independent modules: a state prediction module that predicts the next state based on current state and control inputs, a measurement update module that corrects predictions using sensor measurements, and a gain calculation module that computes optimal weighting. This modular segmentation allows each module to perform a specific function with simple structure, while the combination achieves high estimation accuracy for the state vector.
Solution Approach 2:
The observer employs dynamic gain scheduling where the estimation gain is not fixed but adapts based on operating conditions such as vehicle speed. The gain matrix is recalculated in real-time based on the difference between predicted and actual measurements, allowing the observer to maintain optimal performance across varying operational scenarios without increasing structural complexity.
2Measurement precision
If the observer is optimized for a specific speed range, then the measurement precision improves for that range, but the adaptability to other speed ranges deteriorates
Solution Approach 1:
The observer implements dynamic adaptation by continuously adjusting the estimation gain matrix based on current operating conditions, particularly vehicle speed. The gain is calculated adaptively using the innovation (difference between predicted and actual measurements) and system covariance matrices, allowing the observer to maintain optimal accuracy across the entire speed range from 0 to 200 km/h without requiring separate optimized observers for different speed intervals.
Solution Approach 2:
The observer changes its operational parameters (estimation gain matrix) based on the operating regime. By monitoring system state variables such as vehicle speed and calculating the innovation covariance, the observer automatically adjusts its gain parameters to match current conditions, enabling it to achieve high measurement precision whether the vehicle is moving at low speeds, high speeds, or undergoing rapid acceleration.
3Measurement precision
If real-time adaptation of estimation gain is implemented, then the measurement precision across varying speeds improves, but the computational complexity increases
Solution Approach 1:
The computational task is segmented into distinct calculation stages: prediction of system state, calculation of prediction covariance, computation of innovation (measurement residual), calculation of innovation covariance, and finally gain matrix computation. Each stage processes only the necessary data for that specific purpose, avoiding redundant calculations and reducing overall computational complexity while maintaining real-time adaptation capability.
Solution Approach 2:
The observer implements a balanced level of adaptation complexity by calculating the full gain matrix only when necessary (based on the magnitude of innovation and covariance values) rather than continuously updating all parameters at maximum precision. This partial action approach maintains measurement precision across speed ranges while avoiding excessive computational burden during steady-state operation.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
In order to generate, in real time, an estimated state vector (je) tracking the path of a vehicle (1) moving at a current speed (va), from a control (u) and a current measurement vector (y), the observation device (2) comprises a first module (4) that calculates, in real time, the estimated state vector (je) tracking the path and an estimated measurement vector (y). A second module (5) adjusts, in real time, a temporal variation (x) of the estimated state vector (x) in order to reduce a difference between said current measurement vector (y) and the estimated measurement vector (y) by multiplying the difference by an estimation gain matrix (Lc) calculated by weighted summation, in real time, of a first estimation gain matrix (Lmin) optimised for a first speed (vmin) less than or equal to the current speed (va) and a second estimation gain matrix (Lmax) optimised for a second speed (vmax) greater than or equal to the current speed (va).