Robot Control

Robot Control

The mechatronics development and hardware design activity of the lab is harmonized with a parallel activity on robot control with key focus on the study and implementation of loco-manipulation planning and control principles of humanoids and leg robots as well as on their teleoperation control. The aim of this activity is to develop versatile real-time loco-manipulation control tools by exploring modern optimization techniques and impedance modulation control to realize whole body motion and interaction skills that demonstrate enhanced adaptability permitting robots to execute complex whole body motions while adapting accordingly to real world interaction conditions and uncertainties.


 

Whole test 2 gallery image

Whole test 2 gallery Video

Video Gallery

Divider


Whole Body Motion Control

Whole Body Motion Control

The objective of this activity is to exploit and further develop recent trends in whole-body loco-manipulation technologies, to achieve optimal, coordinated motions of all the robot motors in response to complex, multi-task objectives that naturally arise when dealing with sophisticated robotic platforms, where the trade-off between possibly conflicting goals related to e.g. (i) equilibrium, (ii) loco-manipulation, (iii) collision avoidance, and (iv) the robot physical limits must be solved online as the human operator feeds high-level commands to the robot control system. We focus on hierarchical strategies so that conflicts are sorted out according to a specified hard priority assignment. We study and propose novel formulations that permit to fit the whole-body control paradigm to the hybrid wheeled-legged nature of our CENTAURO platform, allowing to fully coordinate wheeled and articulated motions. Furthermore, we develop force optimization based strategies in order to implement compliant behaviors for floating-base robots.

Publications_WholeBodyControl

Related Publications

A Laurenzi, EM Hoffman, MP Polverini, NG Tsagarakis, An augmented kinematic model for the Cartesian control of the hybrid wheeled-legged quadrupedal robot CENTAURO, IEEE Robotics and Automation Letters 5 (2), pp 508-515, 2019.

EM Hoffman, A Rocchi, A Laurenzi, NG Tsagarakis, Robot control for dummies: Insights and examples using opensot, 2017 IEEE-RAS 17th International Conference on Humanoid Robotics, pp 736-741, 2017.

Multi Contact Loco-Manipulation Planning

Multi Contact Loco-Manipulation Planning

Parallel to our activities on real time control, we also investigate motion planning approaches. As opposed to real time control, where appropriate actions are computed on the basis of the current state of the robot, planning approaches provide the robot with a sense about the time and/or space evolution of a task. We investigate both trajectory optimization based strategies, where the time evolution of the robot behavior with respect to a given task is optimized using gradient-based techniques, and sampling based strategies that explore the geometry of the surrounding environment. We propose and explore novel formulations with the aim to fully exploit the complexity of our multi-limbed robots, with special focus on the ability to plan for the contacts with the environment that best suit the given tasks.

Publications_MultipleContact

Related Publications

MP Polverini, A Laurenzi, EM Hoffman, F Ruscelli, NG Tsagarakis, Multi-contact heavy object pushing with a centaur-type humanoid robot: Planning and control for a real demonstrator, IEEE Robotics and Automation Letters 5 (2), 859-866, 2020.

F Ruscelli, MP Polverini, A Laurenzi, EM Hoffman, NG Tsagarakis, A Multi-Contact Motion Planning and Control Strategy for Physical Interaction Tasks Using a Humanoid Robot, IEEE/RSJ International Conference on Intelligent Robots and Systems, pp 3869-3876, 2021.

Locomotion and Balancing

Locomotion and Balancing

We study locomotion and balancing behaviors targeted to both our quadrupedal and bipedal robotic platforms. Concerning the quadrupedal type, we investigate the use of proprioceptive information such as torque sensing at the joint level in order to autonomously tune a walking gait with respect to varying external conditions such as the payload carried by the robot, as well as its location in the robot body. Bio-inspired joint impedance scheduling behaviors are also developed in order to actively adapt the robot stiffness level to each gait phase (e.g. stance, swing, contact searching), as well as the uncertainty of the perceived robot environment. Furthermore, momentum control strategies exploiting the upper body degrees of freedom can be beneficial in order to increase the robot balancing skills. Research on bipedal locomotion focuses instead on the transition from quasi-static stability principles to less conservative and challenging formulations that allow for phases of free-fall, thus more closely resembling the way humans walk.

Publications_Locomotion

Related Publications

X Zhao, Y You, A Laurenzi, N Kashiri, NG Tsagarakis, Locomotion Adaptation in Heavy Payload Transportation Tasks with the Quadruped Robot CENTAURO, IEEE International Conference on Robotics and Automation, 2021.

C Zhou, X Wang, Z Li, NG Tsagarakis, Overview of gait synthesis for the humanoid COMAN, Journal of Bionic Engineering 14 (1), 15-25, 2017.

C Zhou, Z Li, X Wang, NG Tsagarakis, D Caldwell, Stabilization of Bipedal Walking Based on Compliance Control, Autonomous Robots 40 (6), 1041–1057, 2016.

Teleoperation

Teleoperation

We study robot teleoperation through the use of shared control principles between the robot and the human operator. We investigate Whole-body teleoperation principles of the robotic system by blending the output of robust local controller at the slave side with reference inputs provided by the operators. Whole-body level, local controllers are responsible for regulating critical robotic states which are fed back to the operator through the development of wearable haptic devices based on tactile feedback and novel actuation systems to permit faster and more efficient manoeuvring of the robot under the command of the human operator.