Trajectory forecasting with Transformer networks

[Available]

Introduction

Trajectory forecasting is the task of predicting future objects (or people’s) motion given past trajectories. It is useful for several applications which span from surveillance to autonomous driving [1]. This task is challenging for several reasons: in fact, the future motion depends on interactions among objects and interactions of the object with the scene, which might not be easy to model. In addition, modeling of complex temporal dependencies is required. Using a deep learning model is possible to tackle this problem and predict object’s or people’s future motion in a data-driven fashion. More recently, architectures based on Transformer have been proposed as a solution to trajectories forecasting. Transformers are networks based on attention mechanisms and represent the state-of-the-art in sequence modeling tasks like machine translation and language understanding [2]. Variants of the Transformers are nowadays employed to get more accurate results in the context of trajectory forecasting [3, 4, 5, 6].

The goal of this thesis is to study the recent advances in the field of trajectory forecasting, with particular emphasis on the use of Transformers architectures, and to apply these models to a real problem.

Planned Activities

  1. Initial research on history and state-of-the-art models for trajectory forecasting;
  2. Research on Transformers architectures and how they are applied to trajectory forecasting;
  3. Implementation of state-of-the-art models for trajectory forecasting;
  4. Application of Transformer-based trajectory forecasting algorithms to an industrial problem

Who we’re looking for

Students that are about to get their Master Degree in: computer science, computer engineering, mechatronic engineering, mathematical engineering, mathematics, physics, informatics.

Required skills: 

  1. Proficiency in at least one programming language (Python, C++), Python is preferred;
  2. Basic knowledge of machine learning and Deep Learning (CNN, RNN) algorithms;
  3. Basic knowledge of one of these Deep Learning frameworks (Tensorflow, Pytorch) 
  4. Good mathematical and analytical skills

Duration of this Projects: 6-8 months

Check these links before moving on

Contact Us

Directly by email to: [email protected]

By LinkedIn: linkedin.com/in/cannavò-sonia-66a95467