【摘要】 针对复杂场景中无人机动态预测行人轨迹具有挑战性，提出了一种新的Vanilla-GAN无人机预测行人轨迹方法。由于GAN非常容易受到模式崩溃和梯度下降的影响，Vanilla-GAN方法可以显著改进多模式行人轨迹预测，以避免这些问题。对真实数据和合成数据进行实验，在不同的轨迹数据集上评估不同方法在保持预测分布模式方面的性能，表明该方法能产生更多样点，并保持预测分布。更多还原

【Abstract】 Aiming at the challenge of UAV dynamic prediction of pedestrian trajectories in complex scenes,a new Vanilla-GAN UAV method is proposed to predict pedestrian trajectories. Since GANs are very susceptible to mode collapse and gradient descent,the Vanilla-GAN method can significantly improve multimodal pedestrian trajectory prediction to avoid these problems. Experiments on real and synthetic data show that the method can generate more samples and maintain the predicted distribution.In particular,we design a real trajectory dataset that can be used to evaluate the performance of different methods in preserving the predicted distribution patterns.更多还原