Multi-pedestrian tracking and intention estimation are critical for enhancing the safety and reliability of autonomous vehicles (AVs) in complex driving environments, as these technologies enable AVs to predict and respond to pedestrian behavior more effectively, thereby reducing the likelihood of accidents. This paper proposes an algorithm to model how multiple pedestrians behave and move depending on their environment map. A method that integrates a Gaussian Mixture Probability Hypothesis Density Filter (GMPHD) with the Generalized Potential Field Approach (GPFA) to concurrently track and predict the movements of multiple pedestrians within a short time frame is presented. The algorithm’s reliability was validated through a comprehensive statistical analysis conducted over a substantial dataset of pedestrian movements, demonstrating the model’s reliability and accuracy for tracking and predicting multiple pedestrians in an environment where multiple goals are possible. We aim to determine whether the algorithm can consistently predict pedestrian behavior in dynamic and complex environments, ensuring its applicability in real-world autonomous driving situations.