CrowdNav is a project designed to enhance robot motion planning in environments with dense pedestrian traffic. By integrating Social-LSTM and Nonlinear Model Predictive Control (NMPC), this project offers a robust framework for predicting pedestrian trajectories and dynamically adjusting robot paths to navigate safely and efficiently in crowded spaces.
- Social-LSTM Trajectory Prediction: A deep learning model for accurately predicting pedestrian movements based on social interactions and observed behaviors.
- Nonlinear Model Predictive Control (NMPC): Real-time optimization of the robot's path to avoid collisions and adhere to social norms in dynamic environments.
- Simulation and Real-World Testing: The project includes both simulation and real-world implementations to validate the effectiveness of the proposed framework in various scenarios.
main/: Includes code and configuration for real-world experiments.simulation/: Contains all simulation-related code and resources.
Special thanks to Akin for his guidance and support, and to Valerio, Karim, and all my fellow students who assisted in the real-world experiments.