Seven SLMC papers accepted at IROS 2020
International Conference on Intelligent Robotics and Systems - October 2020
Congratulations to the SLMC group who have had seven papers accepted at The International Conference on Intelligent Robotics and Systems IROS 2020 conference.
The conference is scheduled to take place in Las Vegas from October 25th - 29th.
'Optimizing Dynamic Trajectories for Robustness to Disturbances Using Polytopic Projections'. This paper presents a novel formulation to optimize dynamic trajectories robust to disturbance forces and uncertain payloads. We exploit the structure induced by the kinematics of the robot to formulate a non-trivial nonlinear programming problem that would otherwise be intractable.
Authors: Henrique Ferrolho, Wolfgang Merkt, Vladimir Ivan, Wouter Wolfslag, Sethu Vijayakumar
'Multi-mode Trajectory Optimization for Impact-aware Manipulation'. This paper proposes an impact-aware multi-mode trajectory optimization (TO) method that comprises both hybrid dynamics and hybrid control in a coherent fashion, while impacts are considered through an explicit contact force transmission model. This enables robot manipulators to halt objects with large momentum, e.g. mass of 20kg travelling at speed of 0.88 m/s.
Authors: Theodoros Stouraitis, Lei Yan, João Moura, Michael Gienger, Sethu Vijayakumar
'Online Dynamic Motion Planning and Control for Wheeled Biped Robots'. Wheeled-legged robots combine the efficiency of wheeled robots when driving on suitably flat surfaces and versatility of legged robots when stepping over or around obstacles. This paper introduces a planning and control framework to realise dynamic locomotion for wheeled biped robots. We proposethe Cart-Linear Inverted Pendulum Model (Cart-LIPM) as atemplate model for the rolling motion and the under-actuatedLIPM for contact changes while walking. Combining the two, online dynamic hybrid locomotion has been demonstrated on wheeled biped robots.
Authors: Songyan Xin, Sethu Vijayakumar
‘Automatic Gait Pattern Selection for Legged Robots’. This paper presents a framework that automates gait pattern selection for legged robots, without requiring heuristics or human intuition. More specifically, the framework employs Mixed-integer Nonlinear Programming (MINLP) and a neural network model to create a mapping from task and environment specifications to optimal gait pattern selection. We have tested with a 2D half-cheetah model and the 3D quadruped robot ANYmal. Additionally, we discover qualitative similarities between state-space trajectories for a given gait pattern, which allows us to design initial guess to warm-start trajectory optimisation techniques and achieve online computation rates.
Authors: Jiayi Wang, Iordanis Chatzinikolaidis, Carlos Mastalli, Wouter Wolfslag, Guiyang Xin, Steve Tonneau, Sethu Vijayakumar
'Modeling and Control of a Hybrid Wheeled Jumping Robot'. This paper proposes a receding-horizon approach for the control of a jumping two-wheel robot in simulation. This is done using a proposed template model for its dynamics.
Authors: Traiko Dinev, Songyang Xin, Sethu Vijayakumar
'Optimisation of Body-ground Contact for Augmenting Whole-Body Loco-manipulation of Quadruped Robots'. In this paper we enhance the robustness and manipulation capabilities of quadruped robots by equipping them with prongs: small legs rigidly attached to the body. We analyze the effect of the prongs on robustness, and demonstrate pushing obstacles and picking up a box with the legs of the robot.
Authors: Wouter J Wolfslag, Christopher McGreavy, Guiyang Xin, Carlo Tiseo, Sethu Vijayakumar, Zhibin Li
And in collaboration with CSIC-UPC and NASA-JPL:
'Squash-box feasibility drive differential dynamic programming'. To tackle the limitations of squashing methods in optimal control, we propose a penalization method that combines quadratic barriers and squashing functions and analyze it against a few robotics tasks.
Authors: Josep Marti-Saumell, Joan Sola, Carlos Mastalli and Angel Santamaria-Navarro