Vehicle Behavior Prediction Using Recursive Object Interaction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle behavior prediction methods can only predict situations in the short term, limiting their effectiveness in complex environments.
Innovation Solution
A behavior prediction method and apparatus that uses a microcomputer with integrated sensors and map information to predict the behavior of objects around a vehicle by repeatedly extracting and predicting the behavior of secondary objects, allowing for long-term situation prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Duration of action of moving object
If behavior prediction is based only on driving characteristic of a single vehicle, then prediction processing is simple, but prediction time horizon is limited to short term
Solution Approach 1:
The prediction system segments the traffic environment into multiple hierarchical levels: the host vehicle, surrounding objects (other vehicles, pedestrians, cyclists), and their respective driving characteristics. Each object's behavior is predicted independently based on its own driving characteristic, and these predictions are integrated to form the overall situation prediction. This segmentation allows long-term prediction by breaking down the complex multi-vehicle interaction into manageable individual predictions.
2Measurement precision
If multiple objects are extracted and predicted repeatedly, then long-term prediction accuracy is improved, but operation load increases
Solution Approach 1:
The system performs preliminary extraction of objects surrounding the host vehicle before the main prediction process. By identifying and extracting relevant objects (other vehicles, pedestrians, cyclists) in advance based on their spatial relationship to the host vehicle, the system prepares the data structure needed for repeated prediction cycles. This preliminary action reduces operation load during the actual prediction by having objects pre-identified and organized for efficient processing.
Solution Approach 2:
The prediction system employs periodic action by repeatedly extracting objects and predicting their behaviors at different time steps. The microcomputer periodically performs the extraction and prediction cycle, updating the situation prediction at regular intervals. This periodic repetition allows the system to track long-term behavior patterns while managing computational load through structured, rhythmic processing cycles rather than continuous computation.
3Speed
If driving characteristic is determined from detection information, then real-time prediction is possible, but long-term behavior prediction is insufficient
Solution Approach 1:
The system uses feedback by determining driving characteristics from real-time detection information obtained from sensors (cameras, radar, LIDAR). The detection information about surrounding objects' positions, speeds, and trajectories feeds into the driving characteristic determination, which then informs the behavior prediction. This feedback loop enables the system to adapt predictions to current conditions while extending the prediction horizon by considering how current driving characteristics will evolve over time based on extracted object patterns.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A behavior prediction apparatus specifies a first object that affects a behavior of a vehicle from objects present around the vehicle. The behavior prediction apparatus performs a prediction process of extracting a second object that affects a behavior of the first object among a plurality of objects present around the first object and predicting a behavior. The behavior prediction apparatus sets the extracted second object as a new first object, and performs a prediction process of extracting a new second object affecting the behavior of the new first object and predicting the behavior . The behavior prediction apparatus repeats the prediction process by a predetermined number of times. The behavior prediction apparatus predicts the behavior of the first object in the first prediction process based on the behavior of each of the second objects subjected to the prediction process.