A major challenge today is to exploit recently developed ideas in machine vision and signal processing in the field of collaborative robotics where controlled and repetitive scenarios can rarely be assumed. This project aims to understand and analyse the movement of humans in complex, multi-person environments, often resulting in sophisticated interactions with each other and with other objects in the scene.
In particular, the project consists of two main blocks:
1) artificial vision algorithms to capture and understand human collaborative environments through sensors capable of accurately segmenting actions, predicting behaviours and obtaining three-dimensional measurements. Additionally, audio processing algorithms are considered in this block, with which to obtain other types of information complementary to the geometric semantic information and which allow for greater understanding of the actions and greater robustness in the face of missing modality data. In both cases, massive data and information processing strategies are considered, complemented with physical-geometric observations that will provide greater security to the estimations while increasing the interpretability of the algorithms;
2) exploiting the knowledge acquired in 1) to teach collaborative robots through multi-modal demonstration based on image, point clouds, audio and text. The main model is inspired by a causal transformer that can be trained by autoregressive prediction of acquired human movements. The multimodal nature of the data is exploited by predictions in a modality-aligned way, predicting one token for each input token of the same modality, similar to word prediction in language models.


ABOUT IRI:
IRI is a Joint University Research Institute participated by the Spanish National Research Council (CSIC) and the Technical University of Catalonia (UPC) that conducts research in human-centered robotics and automatic control. The institute is a key player in the Spanish robotics and automatic control scenes, and a valued participant in a large number of collaborations including several ongoing European projects.

ABOUT THE POSTDOC POSITION:
The candidate will join a leading research group in Perception & Manipulation at IRI.
- Training : 240 ECTS of training are also offered, which includes research stays in the University of Cordoba, Spain and UC Berkeley, USA
- Contract conditions: Indefinite contract for a Postdoctoral Researcher associated to the Momentum Project of 4 years' duration according to Spanish science law. Gross annual salary: €41,000 - €52,000.
- Supervisors: Prof. Antonio Agudo and Prof. Manuel Jesús Marín
- Start of contract: Before 31 December 2024.

This position is funded by the “Momentum CSIC Programme: Develop Your Digital Talent”, financed by the Secretariat of State for Digitization and Artificial Intelligence – Ministry for Digital Transformation and Public Service, through Red.es, and with funds from the Recovery, Transformation and Resilience Plan (PRTR)

Contacts

Please, send your documentation before 06/09/2024 to: