The research in robotics has been addressing two topics:

• The development of an energy efficient gait without an area of support.
• Robot vision in dynamic scenes without using prior knowledge.

In the former topic, an abstract method for analyzing a gait has been developed. The result is used to create a gait controller for a Nao robot, which is fine-tuned through reinforcement learning. This movie shows the resulting gait on the Nao robot. The gait uses no ankle stiffness and therefore the robot is constantly falling. The absence of ankle stiffness makes it possible to walk on uneven terrains, as is shown in this movie. The gait reduces the energy consumption by 41% on flat terrains. The method that has been used to develop the new gait, can also be used for other robot movements.

In the latter topic, a vision system is developed that identifies objects based on their movements. The method uses tracking of feature points between pairs of images, affine movement models, the EM algorithm, and Bayesian update. The approach is competitive with the best approaches in the field but computationally 10 times more efficient.

Publications

N. Roos, Z. Sun, Explainable Robotics applied to bipedal gait development, BNAIC (2019). [pdf]

Z. Sun, N. Roos, Dynamically stable walk control of biped humanoid on uneven and inclined terrain, Neurocomputing 280 (2018) 111-122. [pdf]

W. Zhao, N. Roos, R. Peeters, 3D motion consistency analysis for segmentation in 2D video projection,  17th international Conference on Computer Analysis of Images and Patterns (CAIP) (2017) 12 pages. [pdf]

Z. Sun, N. Roos, A Controller for Improving Lateral Stability in a Dynamically Stable Gait, Artificial Intelligence Applications and Innovations (AIAI) (2016) 12 pages. [pdf]

W. Zhao, N. Roos, An EM based approach for motion segmentation of video sequence, International Conferences in Central Europe on Computer Graphics, Visualization and Computer Vision (WSCG) (2016) 9 pages. [pdf]

W. Zhao, N. Roos, Motion based segmentation for robot vision using adapted EM algorithm, International Conference on Computer Vision Theory and Applications (VISAPP) (2016) 649 – 656. [pdf]

Z. Sun, N. Roos, An energy efficient dynamic gait for a Nao robot, IEEE International Conference on Autonomous Robot Systems and Competitions (ICARSC) (2014) 267-272. [pdf]

Z. Sun, N. Roos, Dynamic Lateral Stability for an Energy Efficient Gait, BNAIC (2014) 95-102. [pdf]

W. Zhao, H. B. Ammar and N. Roos, Dynamic Object Recognition using Sparse Coded Three-Way Conditional Restricted Boltzmann Machines, BNAIC 2013 (2013) 271-278.

Z. Sun and N. Roos, An energy efficient gait for a Nao robot, BNAIC 2013 (2013) 191-198.

R. Poddighe and N. Roos, A Nao robot paying tic-tac-toe – Comparing alternative methods for Inverse Kinematics, BNAIC 2013 (2013) 144-151.