Attention Mechanism in Robotic Perception


6c12ddb0 f832 4880 863d c9f8a0a27d03.png 500 500

Organization: Autonomous Systems Lab

Involved Host(s): Cadena Cesar

Abstract: The goal of this project is to develop, implement, and evaluate algorithms (data or model driven) that mimic the attention mechanism, of the human brain, in the robotic perception system.

Description: The role of the attention system is of selecting the most relevant information to perform a given task at a given moment [1,2]. This attention system allows to better interpret, and learn from, multi-modal sensory data (e.g., vision, audio, thermal, range). With an embedded attention mechanism inference can be performed by efficiently exploiting the data and computational resources. The development of such a system poses a number of challenges and research endeavors, such as: (i) active sensing for relevant sensory events -- voluntary (top-down); (ii) detecting signals for intentional (“conscious”) processing -- reflexive (bottom-up); (iii) sustaining alertness to process high priority signals -- covert attention; (iv) guiding for long-term delay rewards; and (v) dealing with sensing pitfalls and previously unseen situations -- plasticity and introspection. This project will be tailored to the student's preference out of the different research topics listed above and level of theoretical/practical research. The applicant should express this preference in the email. [1] S.M. Kosslyn, J.M. Shephard, W.L. Thompson. Spatial Processing during Mental Imagery: A Neurofunctional Theory. In Spatial processing in navigation, imagery and perception, pp. 1-15, Springer, Boston, MA. 2007. [2] M.I. Posner, S.E. Petersen. The attention system of the human brain. Annual Review of Neuroscience, vol. 13, n. 1, pp. 25-42, 1990.

Work Packages: The specific work packages will change depending on the topic selected. However, the general structure follows: - Literature Review on Attention. - Problem setup and assumptions. - Proposed solution: exploration and/or adaptation. - Evaluation.

Requirements: - Highly motivated and independent student. - Strong C++ and/or Python coding skills. - Students from outside of D-MAVT (particularly, from D-INFK, D-ITET, D-PHYS, D-MATH, and Neuroscience) are also highly encouraged to apply.

Contact Details: If you are interested, please send your grade transcripts, CV, and a few sentences about your topic of preference and motivation to Cesar Cadena (cesarc@ethz.ch).

Attention Robotics Perception Deep Learning


Labels: Semester Project Master Thesis CLS Student Project (MPG ETH CLS) ETH Organization's Labels (ETHZ)
Topics: Information, Computing and Communication Sciences Engineering and Technology Behavioural and Cognitive Sciences