Vehicle Roadway Model Estimation Using Multi-Hypothesis Recursion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting the road course for vehicles, especially in highly automated and autonomous driving, face challenges in accurately combining sensor data and map data to determine a reliable road model, often resulting in inconsistent descriptions of the vehicle's surroundings.
Innovation Solution
A method using recursive estimation to generate and evaluate multiple hypotheses of the road model based on sensor and map data, with a Kalman filter for real-time processing, allowing for the selection of the most confident hypothesis and updating map data to reflect changes in the road environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple hypotheses are generated and evaluated using recursive estimation, then the reliability of road model prediction is improved, but the computational complexity and processing time increase
Solution Approach 1:
The road model determination is segmented into multiple hypotheses, each representing a different possible road course interpretation. By dividing the problem into discrete hypotheses rather than attempting a single comprehensive model, the system can evaluate multiple possibilities independently and select the most reliable one, improving prediction reliability while managing computational complexity through structured division of the estimation problem.
Solution Approach 2:
The system generates multiple hypotheses beyond what a single deterministic model would produce, using recursive estimation to evaluate each hypothesis with confidence values. This excessive action of creating multiple potential models ensures that the true road course is captured even when sensor and map data are inconsistent, with the confidence-based selection preventing excessive computational burden from evaluating all possible hypotheses equally.
2Measurement precision
If sensor data and map data are combined to predict road course, then the accuracy of road model is improved, but inconsistencies between data sources reduce reliability
Solution Approach 1:
The system changes the parameter representation by generating multiple hypotheses with different parameter associations between sensor data and map data. Each hypothesis represents a different interpretation of how sensor observations relate to map features, allowing the system to accommodate inconsistencies in the data by selecting the hypothesis with the most consistent parameter associations, thereby maintaining both accuracy and reliability.
Solution Approach 2:
The recursive estimation process incorporates feedback by using confidence values from previous recursion steps to inform current hypothesis evaluation. When inconsistencies between sensor and map data are detected, the feedback mechanism adjusts the confidence values and can trigger map data corrections, allowing the system to learn from data inconsistencies and improve reliability over time while maintaining accurate road models.
3Measurement precision
If map data is corrected based on selected road models, then the accuracy of future predictions is improved, but the system complexity increases
Solution Approach 1:
The system performs preliminary actions by correcting map data based on selected road models before future predictions are needed. By proactively updating map data when inconsistencies are detected, the system prepares more accurate reference information for subsequent hypothesis evaluations, improving future prediction accuracy without adding complexity to the core estimation algorithm itself.
Solution Approach 2:
The system provides self-service by automatically correcting its own map data using the selected road models from recursive estimation. This self-correction mechanism allows the system to maintain and improve its own accuracy without external intervention, reducing the need for complex external calibration systems while improving future prediction accuracy through autonomous map data updates.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the reliability and accuracy of road model prediction, enabling safe and flexible automated driving by continuously updating and refining the road model, even in the presence of short-term changes like construction sites, and reduces the need for high-resolution maps.
Implementation Method 1
Method for determining a road model for surroundings of a vehicle by means of recursive estimation
Implementation Method 2
Acquisition of map data which cartographically characterize the surroundings of the vehicle; Generation of a plurality of different hypotheses for the road model to be determined according to a parameterized state function
Data Source
Figure 1
Figure 2
AI summary
The invention relates to a method and to a device for determining a roadway model (12) by means of recursive estimation and a vehicle (14). The recursion process has the following steps: sensory acquisition of observation data (2, 3) which characterized the surroundings of the vehicle or the movement thereof; acquisition of map data (4) which characterize the surroundings of the vehicle; generation of a multiplicity of different hypotheses (9-11) for the roadway model to be determined, according to a parameterised status function (5A), wherein each of the hypotheses (9-11) represents a possible roadway model for the current recursion step and is characterized by a different Association (6a, 6b), generated by means of corresponding parameterisation, of the observation data (2, 3) and map data (4) which respectively serve as input variables of the state function (5a, 5b); estimating a confidence value relating to a predetermined confidence measure for each of the hypotheses (9-11) of the current recursion step by means of a recursive estimation method using at least one hypothesis (7, 8) from the preceding recursion step as an input variable; selecting, on the basis of the estimated confidence values, one of the hypotheses (9-11) as a roadway model (12) for the surroundings of the vehicle (14) for the current recursion step.