PhD position in robot learning and adaptive control

Idiap Research Institute
September 07 2017
Position Type
Full Time
Organization Type

The Robot Learning & Interaction Group has an open PhD position within the new MEMMO (Memory of Motion) ICT-H2020 European project, in collaboration with 9 other partners (LAAS-CNRS, University of Edinburgh, Max Planck Institute Tuebingen, University of Oxford, PAL Robotics, Airbus Group, Wandercraft, CMPR Pionsat and Costain Group). The project will start in early 2018. The PhD Student will work in tight collaborations with the other partners in the project. 

The ideal PhD candidate should hold a MS degree in computer science, engineering, physics or applied mathematics, with a background in linear algebra, statistics, optimization, signal processing, control and programming. The positions are for 4 years, provided successful progress, and should lead to a dissertation. The selected candidates will become doctoral students at EPFL provided acceptance by the Doctoral School at EPFL ( Annual gross salary ranges from 47,000 CHF (first year) to 50,000 CHF (last year). 

Interested candidates should submit a cover letter, a detailed CV, and the names of three references (or recommendation letters) through the Idiap online recruitment system. 

Interviews will start on October 16, 2017. Late applications will be treated depending on whether the positions have been filled or not. 

MEMMO project description: 

Based on optimal-control theory, the goal of MEMMO is to develop a unified yet tractable approach to motion generation for complex robots with arms and legs. The approach relies on three innovative components: 1) a massive amount of pre-computed optimal motions are generated offline and compressed into a "memory of motion"; 2) these trajectories are recovered during execution and adapted to new situations with real-time model predictive control, with generalization to dynamically changing environments; and 3) available sensor modalities (vision, inertial, haptic) are exploited for feedback control which goes beyond the basic robot state with a focus on robust and adaptive behavior. 

To demonstrate the generality of the approach, MEMMO is organized around 3 applications designed by the end-user partners of the project: 1) a humanoid robot performing advanced locomotion and industrial tooling tasks for aircraft assembly; 2) an advanced exoskeleton paired with a paraplegic patient demonstrating dynamic walking on flat floor, slopes and stairs in a rehabilitation center; and 3) a quadruped robot performing inspection tasks in a construction site. 

In MEMMO, Idiap is responsible of the research aspects related to the representation and encoding of movements. The objective is to compress motion data for fast recognition and adaptive motion synthesis. This will be achieved by extracting invariant structures in a probabilistic form that can be used to generalize movements to new situations (new environments, new contexts, new initial conditions). Models will be developed to facilitate integration between learning and control, with trajectory distributions (incl. the underlying cost functions) that are adapted to the current situation and can be used to quickly generate trajectory samples for further optimization. 

This job comes from a partnership with Science Magazine and Euraxess