Objective:

Motion planning is one of the key research topics in robotics research and, recently, remarkable advancement has been done on the motion planning for humanoid robots, robotic hands, and mobile robots etc. On the other hand, the research on humanoid robots has been extensively done to realize the robot that can work instead of a human. Since the humanoid robots have complex kinematics/dynamics, their motion planning is getting more and more important. This workshop aims to survey the recent advancement on the theoretical/practical aspects of the humanoid motion planning. The main areas of interest of this workshop include but are not limited to:

 

- Whole body motion planning

- Task planning

- Biped gait planning

- Foot step planning

- Sensor feedback for motion planning

 

Organizers:

Kensuke Harada (AIST), Eiichi Yoshida (AIST), and Kazuhito Yokoi (AIST)

 

 

Program:

 

Please click the title to download the workshop materials. Here, each material is locked by the password. The registrants of the workshop are required to send an e-mail to Kensuke Harada (kensuke.harada@aist.go.jp). After confirming the registration, we will provide the password.

 

(1) 9:00 – 9:30

Towards Dynamic Balance Control of Humanoid Robots by using CoM and ZMP

P. Fraisse1, S. Cotton1, A. Murray2, F. Pierrot1

1LIRMM - UMR 5506 Université Montpellier II / CNRS, France

2University of Dayton, USA

 

We propose to introduce an original method to control the location of CoM and ZMP for humanoid robots. This method is initially based on kinematic approach allowing to transform virtually, a tree chain model to a serial chain model for which the CoM position of the tree chain becomes the end effector location of the serial chain. Starting from these developments, we will present a control scheme of the CoM based on classical position control scheme of manipulator arm (serial chain). We demonstrate also that the classical methods dedicated to manipulator arms are easy to use for humanoid robot such, for instance, the computation of the reachable workspace of the CoM. Relying on this method, we propose to extend it to control the location of the ZMP, which corresponds to control the reaction force between the feet and the ground. At last, we will present a hybrid control scheme CoM/CoP inspired by the well-known hybrid position/force control scheme for manipulator arms. Experimental results on humanoid robot HOAP-3 will be presented and commented.

 

(2) 9:30 – 10:00

Whole body motion planning - elements for intelligent systems designs
Michael Gienger and Christian Goerick
Honda Research Institute Europe GmbH, Germany


Controlling the motions of a humanoid robot is a non-trivial task due to the involved number of degrees of freedom of the underlying kinematics and dynamics. Additional constraints arise if the motion control and planning is embedded in an intelligent systems architecture. One of the most crucial questions within this research area is the following: What are appropriate controllers, representations and interfaces in order to command and generate flexible and safe motions efficiently and with a low effort for non-control oriented parts of the overall systems architecture? We will report on our research addressing this question comprising
* task oriented whole body motion control,
* real-time planning using internal prediction and strategy selection,
* robustness issues in movement generation and planning ("task intervals"),
* movement primitives based trajectory planning
* integration of whole body and grasp planning

 

(3)10:00 – 10:30

 

Humanoid motion research at Japanese-French joint lab JRL on HRP-2

 

Eiichi Yoshida, Jean-Paul Laumond, Aberrahmane Kheddar, and Kazuhito Yokoi

 

In AIST/IS ? CNRS/ST2I Joint Japanese-French Robotics Laboratory (JRL), we are conducting research for the objective of increasing robot autonomy particularly via humanoid robot platform in Japan and in France.

 

We present researches activities related to motion planning.

- Whole-body dynamic motion planning: we present how to deal with dynamic for a task that need manipulation and locomotion at the same time. We also address a method for whole-body motion generation, to make a stepping motion to extend its accessible space by a unified framework of generalized inverse kinematics. A method of whole-body collision avoidance to go under obstacle is also under development.

- Manipulation planning

A "pivoting" manipulation of a large object is addressed. We are studying a planning algorithm to arbitrary position and orientation on a plane based on analysis of gsmall-time controllability (STC)h of pivoting operations.

- Contact support point planning

This work deals with the motion planning of a poly-articulated robotic system for which support contacts are allowed to occur between any part of the body and any part of the environment.

 

(*) 10:30 - 10:45  Coffee Break

 

(4) 10:45 - 11:15

Synthesis and Control of Whole-body Behaviors in Humanoid Systems

Luis Sentis

Stanford University, USA

 

A great challenge for humanoid systems is their ability to carry out human-like manipulation and locomotion tasks in dynamic environments. There is a strong need to develop new control architectures that can provide advanced task capabilities as well as high flexibility to operate with high-level planners. In this talk I will present a control methodology for the synthesis and control of realtime whole-body behaviors in humanoid systems as well as several related applications. I will first describe a unified control framework that integrates motion control, contact, and handling of motion constraints in the form of torque-based motion primitives. I will then describe our recent work on control of feet centers of pressure and internal forces between supporting limbs for optimal contact stability. I will proceed by introducing a new approach for online Zero Moment Point pattern generation to synthesize human-like walking behaviors. Finally, I will describe control methods for motion reconstruction by direct control of marker data and their integration into whole-body controllers.

 

(5) 11:15 – 11:45

Planning and Learning Humanoid Motions

Marcelo Kallmann
 University of California, Merced

 

Motion planning for humanoid agents requires algorithms beyond the traditional search for a collision-free path between initial and goal configurations. I will present in this talk an overview of my recent research towards higher level motion planning, learning and cognition. Two main topics will be presented: (1) planning approaches for handling the synchronization (concurrency and sequencing) between primitive motion controllers in order to achieve complex whole-body coordinated locomotion and manipulation, and (2) first attempts toward learning situated tasks with the Attractor-Guided Planner (AGP). The approach taken by AGP is to maintain a task database of previous solutions indexed by both task and environment features. Suitable previous solutions are then selected for guiding the planning of a new task. Examples of applications to computer animation will also be presented.

 

(6) 11:45 – 12:15

Applying Motion Planning Technology to the Real Environment
Koichi Nishiwaki1, Joel Chestnutt2, Mike Stilman2, Satoshi Kagami1, James Kuffner2
1AIST, Japan
2Carnegie Mellon University, USA
 

Motion planning technology is indispensable for large-DOF humanoids to be operated in complicated environments. We have been working on developing footstep planning algorithm that works well in real environments. Vision based environment recognition technology and online walking control system suitable for footstep planning are also developed. We will overview the history of our development to describe what were the difficult problems. A mixed-reality robot experiment environment that helps our development is then presented. Recognition and planning results are overlayed to the real environment so that we can verify those results in the real environment during the experiments. Virtual obstacles can also be introduced instead of using recognition results by overlaying them to the real environments. Other motion planning topics, such as, whole-body reaching system and walking control system that can walk through narrow passages, are also presented as examples of realization on humanoids.

 

(*) 12:15 – 13:30  Lunch Break

 

(7) 13:30 – 14:00

Vision-guided Behavior Control in Motion Planning Humanoids
 Kei Okada, Mitsuharu Kojima, Satoru Tokutsu, Toshiaki Maki, Yuto Mori, Asuka Kadowaki, Masayuki Inaba
The University of Tokyo, Japan

We present a vision based humanoid behavior control for adapting the planned motion to the real world environment. Motion planning is powerful tool for a humanoid performing daily environment tasks. However, for a robot in a real world, perception system to reduce uncertainty is one of the important components to be considered. In this presentation, we presented our knowledge based humanoid robot system that integrates both motion planner and visual recognition subsystem as a unified process. Our system architecture shares the manipulation and the visual features knowledge between planning and perception subsystems in order to guide visual attention relevant to the planned or planning task. Two vision-guided behavior controls are presented. One is visual self/object localization through multi-cue integrated object detection with the visual feature knowledge. Another one is visual behavior verification based on attention control using the manipulation knowledge. Finally, tea serving performances by humanoid robots are shown to demonstrate the feasibility of developed system.

 

(8) 14:00 – 14:30

Navigation Planning for Digital Actors
Julien Pettre

INRIA, IRISA, Campus de Beaulieu, 35042 Rennes, France


Animating Digital Actors is a difficult and time-consuming task. It is possible to ease this task by giving an autonomy of motion to them using, for example,  motion planning techniques. Similarities between Digital Actor models and Robot models, especially Humanoid Robots, allow the use of motion planning techniques, mainly developed by the Robotics field, to reach such an objective. However, a direct application of these techniques is not possible as some new elements must be taken into account such as, among others, the need for realism of motion (or believability) or the need for interactivity. In this talk, we will present some recent motion planning techniques adapted to the Computer Animation needs, addressing problems ranging from a single actor navigating in a scene made of 3D obstacles to the one of a crowd made of thousands of characters populating a large-scale virtual environment.

 

(9) 14:30 – 15:00

Motion Planning for Walking Pattern Generation of Humanoid Robots

Kensuke Harada

AIST, Japan

In this research, we plan the collision free motion for walking pattern generation of a humanoid robot. Our motion planner can  take into account several features of the walking pattern generator.  We first run the walking pattern generator by considering the contact  wrench applied to the robot and monitor the collision  among the links and the environments.  Then, we plan the collision free motion for the period of time causing  the collision. In our motion planner, we can consider the constraint condition which are the functions of time. Also, for keeping balance of the robot, we plan the motion with keeping the horizontal position of the COG as well as the position/orientation of the feet/hand. The effectiveness of the proposed method is confirmed by simulation and experiment.

 

(10)  15:00 – 15:30

Efficient Motion and Grasp Planning for Humanoid Robots

Tamim Asfour, Niko Vahrenkamp, Pedram Azad and Rüdiger Dillmann
University of Karlsruhe (TH), Germany

Motion planning for humanoid robots with many degrees of freedom operating in complex environments is a challenging task. To increase the robotsf autonomy in such environments, a motion planner has to react on dynamic obstacles in reasonable time. To deal with the complexity of the problem we rely on a multiresolutional planning system which takes the different kinematic parts of the body into account and is able to task-dependently combine different planning algorithms in order to reduce the runtime of the planning process.

In this talk we present a multiresolutional motion planner and techniques for efficient motion planning of a humanoid robot acting in a kitchen environment. In addition we present an integrated system for the programming of manipulation and grasping tasks in a household environment. The system incorporates a vision system for the localization and recognition of objects, a path planner for the generation of collision-free trajectories and an offline grasp analyzer that provides the most feasible grasp configurations for the objects encountered in the environment.

 

Contact:

Kensuke Harada, Dr., Eng.

E-mail: kensuke.harada (at) aist.go.jp
Humanoid Research Group, Intelligent Systems Research Institute
National Institute of Advanced Industrial Science and Technology(AIST)
1-1-1 Umezono, Tsukuba, Ibaraki 305-8568, JAPAN
Tel. +81-29-861-5953,  Fax. +81-29-861-5444