Pedestrian Movement Model Learning With Stationary Object Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model parameter learning methods for movement mode models do not adequately consider stationary objects when determining movement trajectories in environments with multiple moving and stationary objects, leading to potential interference.
Innovation Solution
A method that acquires time series data of movement mode parameters, positional relationships, and environment information to learn model parameters for movement mode models, incorporating the extension state of stationary objects, using machine learning algorithms and two-dimensional maps or optical flow data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If mask image is created using only movement trajectory of pedestrians without considering stationary objects, then learning data creation is simplified, but the model parameter learning does not account for stationary objects leading to potential interference
Solution Approach 1:
The patent combines multiple data sources (movement trajectory, positional relationship information, and environment information) into a unified learning data structure. This merging allows the model to simultaneously consider moving pedestrians and stationary objects, resolving the contradiction between data creation simplicity and trajectory accuracy by integrating rather than simplifying the data collection process
Solution Approach 2:
The patent performs preliminary data collection and synchronization of multiple information types before model training. By pre-acquiring and synchronizing environment information with movement trajectories, the system prepares comprehensive learning data in advance, ensuring both accuracy and efficiency during the actual learning process
2Reliability
If time series data from multiple sources is collected and synchronized for learning, then the model can determine movement modes avoiding interference with both moving and stationary objects, but the data processing complexity increases
Solution Approach 1:
The patent introduces a synchronization mechanism that acts as an intermediary to coordinate multiple data sources. This mediator aligns time series data from movement trajectories, positional relationships, and environment information, managing the complexity of multi-source integration while ensuring accurate temporal correspondence for reliable model training
Solution Approach 2:
The patent creates a universal learning data structure that can accommodate multiple types of information (movement modes, positional relationships, environment data) in a unified format. This multi-functional data structure handles diverse information types through a single processing framework, reducing overall system complexity while maintaining comprehensive data representation
Data Source
AI summary
In a learning method, a time series of a movement direction when a reference pedestrian M1 moves to a destination multiple times, a time series of a mask image indicating a positional relationship of nearby pedestrians M2 in a movement direction of the reference pedestrian M1, and a time series of an environment information image 35 are acquired, learning data is created by associating these time series with each other, and a model parameter of a CNN 33a is learned by a back propagation method using the learning data.