Funding

The group is collaborating actively with major research partners. Through previous and current European projects, the group has an extensive collaboration network with partners around Europe.

On-going Projects

B-REAL

Duration: 2020-2022
Funded by: Academy of Finland

B-REAL aims to bridge the reality gap between autonomous learning in real and simulated worlds, where performance and safety are addressed before the real-world interaction.

The Intelligent Robotic’s group role in this project is: 1) address the efficient exploration of the simulated worlds and 2) develop constrained controllers for safety operation in the real-world.

IPALM

Duration: 2019-2022
Funded by: Academy of Finland, European Commission

The purpose of IPALM is to develop methods for the automatic digitization of objects and their physical properties by exploratory manipulations. These methods will be used to build a large collection of object models required for realistic grasping and manipulation experiments in robotics.

Our methods will learn the physical properties essential for perception and grasping simultaneously from different modalities: vision, touch, audio as well as text documents. Category-level prior models will be built by exploiting online available resources. A perception-action-learning loop will then use the robot’s vision, audio, and tactile senses to model instance-level object properties, guided by the more general category-level model. In return, knowledge acquired from a new instance will be used to improve the category-level knowledge.

ASSET

Duration: 2018-2021
Funded by: Academy of Finland

The AI Spider Silk Threading (ASSET) Project goal is to mimic the pultrusion mechanism of spiders using machine learning and robotic micro-manipulation to produce high-toughness artificial silk.

The role of the Intelligent Robotic’s group in this project is the development of data-efficient reinforcement learning algorithms to optimize the silk threading process as well as maximizing the information gain of the artificial silk properties.

ROSE

Duration: 2017-2020
Funded by: Academy of Finland

The ROSE Project – Robots and the Future of Welfare Services – conducts research on how the advance of service robot technologies enables the creation of innovative new products and services and the renewal of welfare services for the aging population.

The Intelligents Robotics group’s role in the project is both the consortium leader as well as the development, integration and assessment of robotic technology for welfare and health services. The Care-O-Bot 4 will be used as dedicated platform for service robotics research.

HBP

Duration: 2018-2020
Funded by: European Commission

The Human Brain Project is a major international effort that studies computational approaches for understanding the brain. We participate in the Neurorobotics sub-project of HBP which looks at how physical embodiment and the brain interact.

Past Projects

Deepen

Duration: 2018-2019
Funded by: Academy of Finland

The goal of the Deepen project was to develop methods for transferring reinforcement learning policies trained in low-fidelity simulations to the physical world. The project was a collaboration between Aalto University and University of Eastern Finland.

During the course of the project, we have worked on sim-to-real transfer of policies trained on low-fidelity images, and on few-shot adaptation of reinforcement learning policies trained using inaccurate simulations of system dynamics.

CCR

Duration: 2017-2019
Funded by: Academy of Finland

Cognitive Cooperative Robots (CCR) project studies the industrial uptake of recent developments in cognitive robotics such as learning from human demonstrations. The project is run in collaboration with the VTT – the national technology agency and Finnish industry.

SEA-SPIDER

Duration: 2015-2018
Funded by: Academy of Finland

The goal of the project SEASPIDER was to enable the learning and execution of autonomous contact tasks, such as assembly or disassembly, in a challenging underwater environment.

For the first time, learning from demonstration was shown to work for in-contact tasks with hydraulic heavy-duty impedance-controlled manipulators capable of wielding payloads of hundreds of kilograms.