Driving Intention Determination via Unified Risk Field Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing driving intention determining methods based on Bayesian networks face challenges in accurately predicting driving intentions due to difficulties in obtaining conditional probability distributions, leading to low accuracy and efficiency, especially when determining intentions for multiple vehicles.
Innovation Solution
A method that constructs a risk field in the driving environment by combining kinetic and potential energy fields from all traffic targets, using a formula that incorporates the status information of each target, including its type, location, velocity, and acceleration, to simplify the process and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If three independent Bayesian network models are established for each another vehicle to determine driving intention, then the determination can be performed for multiple vehicles, but the process becomes complex and determining efficiency is relatively low
Solution Approach 1:
The patent merges the three separate Bayesian network models (destination intention, track intention, velocity intention) into a single integrated model. The model uses a unified probability calculation formula that simultaneously processes multiple intention types by introducing an intention type indicator variable, thereby reducing process complexity while maintaining the capability to determine driving intentions for multiple vehicles.
2Measurement precision
If three independent Bayesian network models are established for each another vehicle to determine driving intention, then comprehensive intention prediction can be achieved, but determining efficiency is relatively low
Solution Approach 1:
The patent combines three separate prediction processes into one unified model that calculates probabilities for all intention types simultaneously. This merging maintains comprehensive prediction capability while significantly improving determining efficiency by eliminating the need to sequentially process three independent models.
Solution Approach 2:
The unified Bayesian network model serves multiple functions by handling destination intention, track intention, and velocity intention prediction within a single framework. The model uses a universal probability calculation formula that can determine any type of driving intention, making the system multi-functional and efficient.
3Measurement precision
If conditional probability distribution is used in Bayesian network model to calculate intention probability, then intention prediction can be performed, but the conditional probability distribution is difficult to obtain accurately
Solution Approach 1:
The patent transforms the difficult-to-obtain conditional probability distribution into a more manageable form by using a unified probability calculation formula with an intention type indicator variable. This parameter change allows the model to calculate probabilities for multiple intention types using a consistent mathematical framework, making the probability distribution more obtainable and accurate.
Data Source
AI summary
A driving intention determining method includes obtaining status information of traffic targets in a driving environment, where the traffic targets include a traffic target in a moving state and traffic targets in a static state, and status information of the traffic targets in a static state includes at least indication information of a traffic sign and road boundary information, determining a risk field of the driving environment based on the status information of the traffic targets in the driving environment, for any other vehicle in the driving environment, determining a driving track of the other vehicle in the driving environment based on the risk field, and determining driving intention of the other vehicle based on the driving track of the other vehicle in the driving environment.