Vehicle Route Prediction Using Dynamic Linear Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting the probable route of a moving motor vehicle are inaccurate, especially during non-stationary driving conditions, due to reliance on map information and assumptions of constant acceleration and yaw rate, leading to imprecise route prognosis and potential safety issues in driver assistance systems.
Innovation Solution
A method using time-related function specifications for rotational and longitudinal driving states, determined through numerical integration of linear models with time derivatives of measured variables, allowing for accurate prediction of vehicle movement without relying on map information, by assuming a linear dynamic model that accounts for transient driving conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If map information and constant acceleration/yaw rate assumptions are used for route prediction, then the prediction method is simple to implement, but the prediction accuracy deteriorates during transient driving conditions
Solution Approach 1:
The patent applies dynamics by transitioning from static assumptions (constant acceleration and yaw rate) to a dynamic prediction model that continuously updates based on measured values. The linear dynamic model adapts to changing driving conditions by incorporating time derivatives and updating weighting factors, enabling accurate route prediction during transient conditions while maintaining computational efficiency.
2Adaptability or versatility
If map information is required for route prediction, then the prediction can be performed with basic sensor data, but the system cannot adapt to changes in road course such as construction sites
Solution Approach 1:
The patent extracts the dependency on map information by developing a prediction method that relies solely on vehicle sensor measurements. By removing the requirement for map data, the system becomes adaptable to any road conditions including construction sites, while maintaining the necessary quantity of information through direct sensor measurements of vehicle dynamics.
Solution Approach 2:
The system performs self-service by using its own sensor measurements to continuously update the prediction model. The weighting factors are automatically adjusted based on measured values and time derivatives, enabling the system to adapt to changing road conditions without external map information or manual intervention.
3Power
If constant acceleration and yaw rate are assumed over the prediction period, then the calculation is computationally simple, but the route prognosis becomes inaccurate during transient driving conditions
Solution Approach 1:
The patent applies parameter changes by transitioning from fixed parameters (constant acceleration and yaw rate) to time-varying parameters. The linear dynamic model incorporates time derivatives and updates weighting factors based on measured values, allowing the parameters to adapt to transient conditions while maintaining computational efficiency through the structured mathematical framework.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The invention relates to a method for predicting the anticipated route (14) of a travelling motor vehicle (1) by numeric integration of a dynamic vehicle model with at least one rotational driving state and at least one longitudinal driving state. The invention further relates to a prediction device for carrying out the method. In order to predict the anticipated route (14) independently of map information and with the highest possible accuracy, even in final stationary travel states of a vehicle, according to the invention time-related functional rules for the rotational driving state (ϕPre(t)) and/or the longitudinal driving state (vPre(t)) are determined , and by integration with this functional rule (ϕPre(t), vPre(t)) values of the respective driving state (PPre, vPre) are predicted at specific times. In this context, the time-related functional rule of the respective driving state is determined in that in each case rotational or longitudinal input variables are determined from measured values for at least two temporal derivations of the respective driving state, and the input variables are related to one another in linear models of an order corresponding to the number of the input variables (ω0, ѽ0, ao, Ɩ0) of the respective driving state (PPre, vPre)using predefined time constants (Τω,1 ,Τνν,2,Ta,1,Τa,2), and a time-related functional rule (ωPre(t), apre(t)) for anticipated predicted values of the respective input variable is obtained from the linear model. The functional rule (ωPre(t), αPre(t)) for anticipated predicted values of the respective input variable is integrated analytically.