Vehicle trajectory prediction and generation using LSTM models and GANs

Abstract

Vehicles’ trajectory prediction is a topic with growing interest in recent years, as there are applications in several domains ranging from autonomous driving to traffic congestion prediction and urban planning. Predicting trajectories starting from Floating Car Data (FCD) is a complex task that comes with different challenges, namely Vehicle to Infrastructure (V2I) interaction, Vehicle to Vehicle (V2V) interaction, multimodality, and generalizability. These challenges, especially, have not been completely explored by state-of-the-art works. In particular, multimodality and generalizability have been neglected the most, and this work attempts to fill this gap by proposing and defining new datasets, metrics, and methods to help understand and predict vehicle trajectories. We propose and compare Deep Learning models based on Long Short-Term Memory and Generative Adversarial Network architectures; in particular, our GAN-3 model can be used to generate multiple predictions in multimodal scenarios. These approaches are evaluated with our newly proposed error metrics N-ADE and N-FDE, which normalize some biases in the standard Average Displacement Error (ADE) and Final Displacement Error (FDE) metrics. Experiments have been conducted using newly collected datasets in four large Italian cities (Rome, Milan, Naples, and Turin), considering different trajectory lengths to analyze error growth over a larger number of time-steps. The results prove that, although LSTM-based models are superior in unimodal scenarios, generative models perform best in those where the effects of multimodality are higher. Space-time and geographical analysis are performed, to prove the suitability of the proposed methodology for real cases and management services.

GeoAI: a review of artificial intelligence approaches for the interpretation of complex geomatics data

“…This process proceeds until the generator development of outputs. GANs have been used to generate artificial images and videos as well as to generate point clouds (Vondrick et al, 2016;Sun et al, 2020;Rossi et al, 2021). Despite exceptional results in supervised learning since the DL developments, collecting enough data to train the models remains a challenge, and some methods have been developed to train models with little or no data.…”

Traffic Swarm Behaviour: Machine Learning and Game Theory in Behaviour Analysis

“…In a typical approach, inputs come from the eight surrounding vehicles, which concept in [32] is extended by the preceding vehicle of the front vehicle resulting in the most reliable behaviour modelling and prediction. Most of the approaches use a full set of input and output data; however, [33] exploited sparse floating car data to predict vehicle trajectories with penetration rate between 2 and 8 percent.…”

“…Through intensive analysis [33] proved that, although LSTM-based models are superior in unimodal scenarios, generative models perform best in those where the effects of multimodality are higher. Luca et al [33] shows that the difference between using an LSTM and a GAN for prediction lies in the objective function.…”

“…Through intensive analysis [33] proved that, although LSTM-based models are superior in unimodal scenarios, generative models perform best in those where the effects of multimodality are higher. Luca et al [33] shows that the difference between using an LSTM and a GAN for prediction lies in the objective function. The LSTM learns an average behaviour because it minimizes the average error between all the predictions, while GANs learn to produce plausible samples corresponding to specific behaviours as a result of competition between generator and discriminator networks.…”

BubblEX: An Explainable Deep Learning Framework for Point-Cloud Classification

“…Considering the 2D counterpart (e.g. image analysis [5] or trajectory data [6]) it can be stated that point cloud processing is one step behind [7]. Indeed, the extraction of semantic information from images has been revolutionised by deep learning techniques, in particular by convolutional neural networks (CNN), which have been shown to outperform other techniques.…”