• The Journal of Korea Robotics Society
• /
• v.14 no.1
• /
• pp.40-49
• /
• 2019

Reinforcement learning has been applied to various problems in robotics. However, it was still hard to train complex robotic manipulation tasks since there is a few models which can be applicable to general tasks. Such general models require a lot of training episodes. In these reasons, deep neural networks which have shown to be good function approximators have not been actively used for robot manipulation task. Recently, some of these challenges are solved by a set of methods, such as Guided Policy Search, which guide or limit search directions while training of a deep neural network based policy model. These frameworks are already applied to a humanoid robot, PR2. However, in robotics, it is not trivial to adjust existing algorithms designed for one robot to another robot. In this paper, we present our implementation of Guided Policy Search to the robotic arms of the Baxter Research Robot. To meet the goals and needs of the project, we build on an existing implementation of Baxter Agent class for the Guided Policy Search algorithm code using the built-in Python interface. This work is expected to play an important role in popularizing robot manipulation reinforcement learning methods on cost-effective robot platforms.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.24 no.5
• /
• pp.548-552
• /
• 2015

In this paper, we describe the quadruped walking-control algorithm of the complete full-size humanoid DARPA Robotics Challenge-HUBO (DRC-HUBO) robot. Although DRC-HUBO is a biped robot, we require a quadruped walking function using two legs and two arms to overcome uneven terrains in the DRC. We design a wave-type quadruped walking pattern as a feedforward control using several walking parameters, and we design zero moment point (ZMP) controllers to maintain stable walking using an inverted pendulum model and an observed-state feedback control scheme. In particular, we propose a switching algorithm for ZMP controllers using supporting value and weighting factors in order to maintain the ZMP control performance during foot switching. Finally, we verify the proposed algorithm by performing quadruped walking experiments using DRC-HUBO.

• Journal of Institute of Control, Robotics and Systems
• /
• v.18 no.1
• /
• pp.28-34
• /
• 2012

Human generally uses three methods to keep balance. One of them is using reactive momentum such as swing an upper body or arms. In this study, we proposed a balancing controller for the reactive momentum method using an inverted pendulum. We simplified a human or a humanoid robot as a two-link inverted pendulum having two edges. In addition, we proposed a distinctive condition for controller transition. If a human is pushed, he has to change a balancing controller from using an ankle torque to using a reactive momentum or changing foot placement. When the balancing controller is changed from using an ankle torque to using a reactive momentum, it is required a proper timing to keep a stability and make smooth movement. In the experiment, the proposed controller and distinctive condition were verified.

• Journal of Institute of Control, Robotics and Systems
• /
• v.6 no.3
• /
• pp.298-308
• /
• 2000

In this paper a new two-arm cooperative assembly(or insertion) algorithm is proposed. As a force-guided control method for the cooperative assembly the adaptive accommodation controller is adopted since it does not require any complicated contact state analysis nor depends of the geometrical complexity of the assembly parts. Also the RMRC(resolved motion rate control) method using a relative jacobian is used to solve inverse kinematics for two manipulators. By using the relative jacobian the two cooperative redundant manipulators can be formed as a new single redundant manipulator. Two arms can perform a variety of insertion tasks by using a relative motion between their end effectors. A force/torque sensing model using an approximated penetration depth calculation a, is developed and used to compute a contact force/torque in the graphic assembly simulation . By using the adaptive accommodation controller and the force/torque sensing model both planar and a spatial cooperative assembly tasks have been successfully executed in the graphic simulation. Finally through a cooperative assembly task experiment using a humanoid robot CENTAUR which inserts a spatially bent pin into a hole its feasibility and applicability of the proposed algorithm verified.