Thesis Self-Supervised Long-Term Trajectory Prediction

Job title: Thesis Self-Supervised Long-Term Trajectory Prediction

Company: Bosch

Job description: Company Description

At Bosch, we shape the future by inventing high-quality technologies and services that spark enthusiasm and enrich people’s lives. Our promise to our associates is rock-solid: we grow together, we enjoy our work, and we inspire each other. Join in and feel the difference.

The Robert Bosch GmbH is looking forward to your application!

Job Description

Trajectory prediction is one of the key elements of the autonomous driving stack. Despite a strong research and industry focus, there are many problems to be solved, such as understanding complex social interactions among different agents, efficiently incorporating rich topological information, predicting multi-modal trajectories, and achieving reliable long-term predictions. In tackling these challenges, Deep Learning (DL) algorithms have shown promising results over classical robotics approaches, especially for the use-case of urban driving.

In long-term prediction (
3 seconds into the future), the action-based self-supervised approach [1] has shown promise by dividing the future interval into equal time-segments and performing prediction segment-wise. It performs environment context prediction and input reconstruction simultaneously over each segment, enabling a richer understanding of the cause-and-effect relationship between actions and features.

  • This topic concerns vehicle trajectory prediction with a focus on long-term prediction with self-supervision.
  • First, you will perform a conceptual analysis of the problem and state-of-the-art methods.
  • Then, you will be tasked with extending and improving the existing self-supervised model. Practical implementation is performed in Python and Pytorch.

Literature:

[1] Janjoš, Faris, Maxim Dolgov, and J. Marius Zöllner. „Self-Supervised Action-Space Prediction for Automated Driving.“ arXiv preprint arXiv:2109.10024 (2021).

[2] Xiao, Yi, et al. „Action-based Representation Learning for Autonomous Driving.“ arXiv preprint arXiv:2008.09417 (2020).

[3] Yuan, Ye, et al. „AgentFormer: Agent-Aware Transformers for Socio-Temporal Multi-Agent Forecasting.“ arXiv preprint arXiv:2103.14023 (2021).

[4] Scibior, Adam, et al. „Imagining The Road Ahead: Multi-Agent Trajectory Prediction via Differentiable Simulation.“ arXiv preprint arXiv:2104.11212 (2021).

Qualifications

  • Education: studies in the field of Computer Science, Electrical Engineering or comparable with a robotics/machine learning focus and with very good grades
  • Personality: motivated to work on an exciting real-world application of machine learning such as autonomous driving
  • Working Practice: ready to learn a lot, in order to dive into a topic at the frontiers of machine learning research and autonomous driving applications and in case of potential own novel contributions, you should be open to publishing them
  • Experience and Know-how: in reading research papers and coding experience for machine learning applications, in Python with pytorch or tensorflow
  • Languages: fluent in English

Additional Information

Start: according to prior agreement
Duration: 6 months

Requirement for this thesis is the enrollment at university. Please attach a motivation letter, your CV, transcript of records, examination regulations, if available published papers and a publicly available code (github repos, etc) and if indicated a valid work and residence permit.

Need further information about the job?
Faris Janjos (Business Department)
+49 711 811 49109

Apply now in just 3 minutes!

Expected salary:

Location: Renningen, Baden-Württemberg

Job date: Fri, 05 Nov 2021 01:41:40 GMT

Apply for the job now! – Submit your Resume

Work and Live in Germany? find jobs and information about visa language and applications here Jobs-in-Germany.net