Generating Realistic and Representative Trajectories with Mobility Behavior Clustering

Area of Science:

• Computer Science
• Artificial Intelligence
• Urban Planning

Background:

• Realistic human movement trajectories are vital for urban planning, transportation, and public health applications.
• Privacy concerns limit access to real-world trajectory data, necessitating the generation of synthetic data.
• Existing deep neural network (DNN) methods for synthetic trajectory generation often overlook crucial human moving behaviors.

Purpose of the Study:

• To develop a novel framework, MBP-GAIL, for synthesizing realistic human trajectories.
• To incorporate moving behavior patterns into the trajectory generation process.
• To improve the accuracy and utility of synthetic trajectory data for simulations.

Main Methods:

• MBP-GAIL utilizes generative adversarial imitation learning (GAIL).
• Recurrent Neural Networks (RNN) are employed to model temporal dependencies in movement.
• The framework integrates stochastic constraints from moving behaviors and spatial constraints.

Main Results:

• MBP-GAIL successfully synthesizes realistic human trajectories that preserve moving behavior patterns.
• The proposed method demonstrates superior performance compared to state-of-the-art trajectory generation techniques.
• The generated trajectories enhance decision-making capabilities in trajectory simulation applications.

Conclusions:

• MBP-GAIL offers a significant advancement in generating realistic synthetic human trajectories.
• Incorporating moving behavior patterns is key to improving trajectory synthesis accuracy.
• This framework has strong potential for applications in urban planning, transportation, and public health simulations.