Robotics Research Highlights

Overview

Activities of the Robotics Group at BISG include comprehensive knowledge on building robotic platforms for indoor, aquatic and terrain environments: localisation, environmental modelling, software architectures and human-robot interaction, and in particular magnetic field localisation and SLAM, dynamic motion planning, winter operability for electric vehicles, autonomous water quality sensing.

Fundamental research questions include:

  • Information representation in robot’s software for Integration of algorithms; adaptive systems for current conditions;
  • Advanced performance for robots involving real time requirements as well as reflex, dynamic control, fast adaptive perception and smooth motion trajectories and obstacle avoidance;
  • Energy and cost effective robot building implying low-cost electronics, multiprocessor embedded systems as well as power consumption optimisation and knowledge, and consuming and storing energy efficiently;
  • Distributed control systems for robotics, calling for ad hoc communication and sensing networks, small embedded nodes for building robots, multi-robot systems (wheeled, arms, and flying robots);
  • Routine learning and using it for task planning and scheduling and related statistics.

Flexible Robot Programming

Modularity enables flexible and reusable design and speeds up prototyping and system development. To achieve this goal, interfaces must be simplified. Decoupling robot control from management of the interfaces towards sensors and actuators, lowers the threshold and allows researcher other than embedded system experts to investigate robotics. The Atomi framework concept (Vallius & Röning 2006) implemented in Qutie robot (Tikanmäki et al. 2007) was developed at BISG before the establishment of robot operating systems such as ROS.

A related spin-off company is Atomia (embedded systems).

Human-Robot Interaction

Multimodal perception exploiting many sensors, external camera network is used for natural human-robot interaction involving detection of emotion and gestures of humans. Tasks include fetching objects on a table. A research problem here is the information extraction: understanding terms in spoken language and dialogue, fetching information from Internet and visual speech synthesising.

This work was part of Minotaurus project. See also the old research topic page.

Mobile Robots

Mobile robots typically operate in unknown environment calling for research on perception, life-time learning, harsh environment.

Mobile robotics are complex real-time systems, hence the research question on how to handle the complexity of building a robotic system. As a reference, humanoid robots include about 100 actuators and more than 100 sensors, thus implying teamwork of mechanics, electronics and software. In addition to complexity, cost is also an issue. For example, the cost of Honda Asimo exceeds one million euros.

Indoor Robotics

Motion and force control involves kinematics of complex wheeled robots, force control of wheeled robots, decentralised control, smooth jerk-limited motion, and formation control. Example: complex delta-robot control.

Indoor navigation based on exploitation of local magnetic field distortion lead to spin-off IndoorAtlas, whose cloud-based technology allows a localisation accuracy of about one to two meters.

See also the old research topic page.

Outdoor Robotics

The aim is control of mobile robots in harsh environment. On this theme, the group participated with success in several international robot competitions: C-Elrob 2007 (second in combined scenario), M-Elrob 2008 (winner in Camp security, fourth in Mule scenarios). C-Elrob 2009 (Oulu, 15-18 June), M-Elrob 2014 in Poland ("best innovative solution" award).

The Robotics Group team participated the European Space Agency's Lunar Robot Challenge 2008 and ranked third with the smallest number of penalties (see ESA's article).

Practising

Day 1 and Day 2&3 at the competition

In addition to UGVs above, UAV-UGV cooperation is studied, namely: environment scouting for UGV, radiation measurement with UAV, ground mapping and 3D reconstruction in real-time. Low cost UAV (about 100 EUR copter with external vision) swarms are used.

The activities on ground robots originated the spin-off Probot, founded 2006. Its expertise is in: modular robotics, easily customisable frames, distributed control nodes, real world applications (logistics, gardening, farming, site survey, environment measurement), modular manufacturing (D3000 robot: PCB testing, 3D printing, small series manufacturing), embedded nodes for distributed control.

The research activities on surface robots originated the spin-off Aquamarine Robots. The aims are cost-efficient sensing strategies in terms of sensing time and modeling accuracy to collect water quality information from the target environment. Additional challenges in the water environment follow from the changing weather conditions and fishing/leisure boats navigating in the same area.

Evolutionary Robotics

Evolutionary Robotics on Lego NXT Platform. In Poikselkä, Vallivaara & Röning (2015) it is shown that the low-cost Lego NXT educational set is indeed adequate for simple experiments in evolutionary robotics. This is demonstrated by an experiment, where an artificial neural network-based controller capable of behaving meaningfully in a Lego sumo wrestling context is evolved on physical Lego NXT robots without the aid of simulation. To the authors’ knowledge this is the first time the non-simulated Lego NXT is used to conduct artificial neural network-based evolutionary robotics.

Education and Collaboration

In the framework of euRathlon EU project, the group organised in Oulu the euRathlon TRADR summer school 2016 and the euRathlon Sherpa summer school 2015. In the 2015 edition, alongside lectures by experts, a cooperative task between aerial and surface vehicles was demonstrated.

Last updated: 22.9.2017