Motion Planning for Industrial Robots
ICRA 2014, June 05
Abstract
Algorithmic motion planning has been actively studied in robotics and related areas for more than three decades. Although there is a rich collection of motion planning algorithms and their applications to CAD/CAM, bioinformatics, and gaming, the use of motion planning techniques on industrial robots has been limited.
Following the success of an international forum on this topic held at iREX 2013 in Tokyo, we are organizing this workshop at ICRA 2014 to address the gap between theory and practice. This workshop is expected to address these gaps and bring together leading academic and industrial researchers/practitioners with varying backgrounds, who can address different aspects of these problems. Furthermore, most of these presenters have also worked on porting the appropriate planning techniques to physical robots.
The workshop program consists of invited talks as well as contributed presentations from researchers in industry and academia. We aim to share their experiences and also point out major open issues.
Workshop Schedule
9:00 a.m. - 10:30 a.m.
Chaired by Dinesh Manocha
| 9:00 a.m. | Motion Planning for Humanoid Robots: Feedback from HRP2-14 Platform Jean-Paul Laumond (LAAS-CNRS, Toulouse) SLIDES |
| 9:30 a.m. | Robot Programming for Assembly Tasks Yukiyasu Domae & Akio Noda (Mitsubishi Electric Corporation) SLIDES |
| 10:00 a.m. | Automated Planning for Supporting Human Robot Collaboration in Assembly Cells Satyandra Gupta (University of Maryland) SLIDES |
10:50 a.m. - 12:30 p.m.
Chaired by Jean-Paul Laumond
| 10:50 a.m. | Overview of IEEE RAS TC Kostas Bekris SLIDES |
| 11:00 a.m. | Many-robot Path Planning & Other Warehouse Robot Challenges Joseph Durham (Kiva Systems) |
| 11:30 a.m. | Motion Planning & Proximity Computations for Industrial Robots: Challenges & Lessons Learned Dinesh Manocha (UNC-Chapel Hill) SLIDES |
| 12:00 p.m. | Control & Motion Planning for Flexible Parts Assembly Kensuke Harada, Juan Rojas & Eiichi Yoshida (AIST, Tsukuba) SLIDES |
2:00 p.m. - 3:30 p.m.
Chaired by Eiichi Yoshida
| 2:00 p.m. | Motion Planning for Industrial Robots using MoveIt! Sachin Chitta (SRI International) SLIDES |
| 2:30 p.m. | When the Travelling Salesman Meets the Piano Mover in a Digital Factory Etienne Ferre (Siemens PLM Software, Toulouse) SLIDES |
| 3:00 p.m. | Integrating Dynamics into Industrial Motion Planning Quang-Cuong Pham (Nanyang Technological University, Singapore) & Huan Liu (Mujin Inc.) SLIDES |
3:50 p.m. - 5:30 p.m.
Chaired by Etienne Ferre
| 3:50 p.m. | Overview of IEEE RAS TC Kensuke Harada |
| 4:00 p.m. | Similar Part Rearrangement With Pebble Graphs Athanasios Krontiris, Rahul Shome, Andrew Dobson, Andrew Kimmel, Isaac Yochelson, Kostas E. Bekris SLIDES |
| 4:30 p.m. | Dual Arm Manipulation using Constraint Based Programming Yuquan Wang, Francisco Viña, Yiannis Karayiannidis, Christian Smith, Peter Ögren SLIDES |
| 5:00 p.m. | Trajectory Planning for Robots: The Challenges of Industrial Considerations Joonyoung Kim, Elizabeth A. Croft SLIDES |
| 5:30 p.m. | Presentation by Mujin Rosen Diankov SLIDES |
Organizers
- Jean-Paul Laumond, LAAS-CNRS, 7 Avenue du Colonel Roche, 31077 Toulouse, France; jpl(at)laas.fr, +33(0)561336347
- Dinesh Manocha, Department of Computer Science, University of North Carolina, Chapel Hill, NC 27599-3175, USA; dm(at)cs.unc.edu, +1(919)962-1749
- Eiichi Yoshida, AIST, Tsukuba Central 2, 1-1-1, Namiki, Ibaraki 205-8568, Japan; e.yoshida(at)aist.go.jp, +81(29)861-7108
Presenters & Abstracts
Sachin Chitta (SRI International)
Title: Motion Planning for Industrial Robots using MoveIt!
Abstract: This talk will present an overview of the MoveIt! motion planning software. MoveIt! is designed to provide a motion planning solution for manipulators and mobile manipulators with tight integration with 3D perception. MoveIt! has found wide application in industry as a software platform for enabling advanced robotics applications like generalized pick and place. MoveIt! has now been ported to over 65 different robots including a wide range of industrial manipulators from all the major vendors. This talk will include a description of the technology available in MoveIt!, use-cases on how MoveIt! is being used and future directions for new applications in industry.
Bio: Sachin Chitta, Ph.D., is associate director of robotics systems and software in the Robotics Program at SRI International. Before joining SRI, Sachin Chitta was at Willow Garage from 2007-2013 and was a core member of the team that developed the PR2 robot and the Robot Operating System (ROS). He initiated and led the development of the MoveIt! and Arm Navigation software platforms to enable advanced manipulation capabilities for any robot. Sachin was a key member of the team that founded Redwood Robotics, a joint venture between Willow Garage, MEKA and SRI. Sachin was a finalist for the 2013 World Technology Award in the IT Software (Individual) category. Sachin was the keynote speaker at ROSCON 2013. His work has won best paper awards at ICAR (2009), ASME IDETC (2008) and been nominated for a best paper award at ICRA 2011. His work on the Open Motion Planning Library (in collaboration with Lydia Kavraki's group at Rice University) won the OSS World Challenge Grand Prize in 2012. Sachin received his Ph.D. from the Grasp Laboratory at the University of Pennsylvania in 2005.
Rosen Diankov (Mujin Inc.)
Abstract: Mujin's technology centers on the Mujin Controller, the world's first commercial motion controller for real-time applications like binpicking and other applications like multi-robot manipulation planning. The Mujin Controller delivers reliable motion planning capabilities to the industrial robot system integrators and big manufacturing companies. It can has a easy-to-use API allowing users to dynamically configure it through HTTP, control robots in real-time, and create complex assembly tasks through its Industrial Task Language (ITL). The Mujin Controller has planned for hundreds of robot types and is already gathering a big customer base in the Japanese industrial robotics market.
In this talk, we will touch on all the key applications Mujin is working on and discuss the opportunities and challenges remaining to robotics researchers to take industrial robotics to the next generation.
Bio: Mujin was co-founded in 2011 by Rosen Diankov, a CMU Robotics Institute graduate and developer of the open-source motion planning framework OpenRAVE, and Issei Takino, a manufacturing business expert. In the following year, the core team grew to include Huan Liu, MIT graduate and Luke Doyle, RPI graduate and web engineering expert. The decision to base Mujin in Japan has given the team a huge advantage to work with the best and most rigorous manufacturing companies in the world. Mujin products are already selling and as a result, Mujin is rapidly growing and looking for extremely motivated engineers that want to make an impact in the robotics industry together.
Joseph Durham (Kiva Systems)
Title: Many-robot Path Planning and Other Warehouse Robot Challenges
Abstract: Kiva’s mobile fulfillment system blends techniques from AI, Controls Systems, Machine Learning, Operations Research and other engineering disciplines into the world’s largest mobile robotic platform. Kiva uses hundreds to thousands of mobile robots to carry inventory shelves around an e-commerce warehouse. Mobile inventory enables the optimization of storage and retrieval of physical inventory using techniques developed for computer data, making order fulfillment fast and efficient. Kiva robots are running in over 50 warehouses on multiple continents for Amazon, Staples, The Gap, and others, with many new sites opening everyyear.
This talk will discuss the Kiva coordinated path planning problem. While planning efficient paths for a single robot on a weighted graph is well understood, coordinating the constant motion of thousands presents many challenges. We will look at the impact on path planning of task selection, traffic density, and the inherent uncertainty introduced by human-robot interaction.
In addition, this talk will describe the role of human “pickers” in fulfilling orders. Advances in manipulators and sensors ask the question of whether the picking of product and packing of orders can be automated. We will discuss the order picking problem domain, what makes humans good at it, and challenges facing robotic manipulation solutions.
Bio: Joseph Durham is a Senior Research Scientist at Kiva Systems, a subsidiary of Amazon.com. He received a Ph.D. in Mechanical Engineering from the University of California at Santa Barbara (UCSB) in 2011 for work on distributed algorithms for teams of robots, and has been working on algorithms for teams of Kiva robots ever since. Previously he earned an M.S. in Mechanical Engineering from UCSB in 2007 and a B.A. inPhysics from Carleton College in 2004. He also worked on path planning for the Stanford University Racing Team autonomous car for the 2005 DARPA Grand Challenge.
Etienne Ferre (Siemens Industry Software)
Title: When the Travelling Salesman Meets the Piano Mover in a Digital Factory
Abstract: Increased complexity of products and manufacturing processes presents manufacturers with time-to-market and asset optimization challenges. Manufacturing engineering teams are expected to deliver flawless product launches while adhering to cost, quality and production launch targets. To meet these challenges, leading manufacturers need to utilize organizational knowledge and 3D models of product designs and manufacturing resources in order to virtually validate their manufacturing processes prior to the start of production.
Manufacturing Process Management Software like Tecnomatix Process Simulate provides a virtual environment to reach optimal solutions. One key tool is the automatic path planner developed by Kineo, a business unit of Siemens PLM Software. The path planner allows to automatically computing collision free robot trajectories while respecting all the specific constraints of the industrial robots. Kineo’s robust advanced path planning makes it possible to create trajectories that are safer and faster than classic robot teaching.
This presentation will illustrate how the path planner is integrated in the manufacturer workflow and point to the constraints of the industrial environment. We will examine the use case of Body-in-White applications where many robots have to weld sheet metal components together while trying to optimize the welding sequences. This use case will highlight the challenges that can be addressed with a powerful path planner combined with Travelling Salesman Problem solver algorithms.
Satyandra Gupta (University of Maryland)
Title: Automated Planning for Supporting Human Robot Collaboration in Assembly Cells
Abstract: Human and robots have complementary strengths in performing assembly operations. Humans are very good at perception tasks in unstructured environments. For example, they are able to recognize and locate a part from a bin of miscellaneous parts. They are also very good at complex manipulation in tight spaces. In contrast, robots are very good at pick and place operations and highly repeatable. Robots can perform tasks at high speeds and still maintain precision. Robots can also operate for long periods of times with showing signs of fatigue. Typically, robots are used in mass production lines. Small batch and custom production operations predominantly use manual assembly lines.
The high labor cost is making it difficult for small and medium manufacturers to remain cost competitive in high wage markets. These manufactures are mainly involved in small batch and custom production. They need to find a way to reduce the labor cost in assembly operations. Purely robotic cells will not be able to provide them the necessary flexibility. Creating hybrid cells where humans and robots can collaborate in close physical proximities is a potential solution. The underlying idea behind such cells is to decompose assembly operations into tasks such that humans and robots can collaborate by performing subtasks that are suitable for them.
This presentation will describe the on-going assembly planning research at the Maryland Robotics Center to realize hybrid assembly cells to enable safe and efficient human and robot collaboration. The following three topics will be covered as a part of this presentation:- We need to be able to automatically generate plans to operate hybrid assembly cells to ensure efficient cell operation. This requires generating feasible assembly sequences and instructions for robots and human operators, respectively. Automated planning poses the following two challenges. First, generating operation plans for complex assemblies is challenging. The complexity can come due to the combinatorial explosion caused by the size of the assembly or the complex paths needed to perform the assembly. Second, generating feasible plans requires accounting for robot and human motion constraints. I will describe algorithms for automatically generating plans for the operation of hybrid cells. It will address both assembly complexity and motion constraints issues.
- The collaboration between humans and robots in the assembly cell will only be practical if human safety can be ensured during the assembly tasks that require collaboration between humans and robots. I will describe different options for real-time monitoring of the state of human operator with respect to the robot and strategies for taking appropriate measures to ensure human safety when the planned move by the robot may compromise the safety of the human operator.
- If the human operator makes an error in selecting the part or placing it correctly, the robot will be unable to correctly perform the task assigned to it. If the error goes undetected, it can lead to a defective product and inefficiencies in the cell operation. In order to ensure smooth and error-free operation of the cell, we will need to monitor the state of the assembly operations in the cell. I will present algorithms to identify and track parts in the cell and automatically generate instructions for taking corrective actions if a human operator deviates from the selected plan. Potential corrective actions include replanning if it is possible to continue the assembly operation from the current state and issuing warning and generating instructions to undo the current task.
Kensuke Harada, Juan Rojas & Eiichi Yoshida (AIST, Japan)
Title: Control and Motion Planning for Flexible Parts Assembly
Abstract: In this talk, we will show some recent developments on
control and motion planning for flexible parts assembly.
We first discuss on snap assembly since most of the plastic
pats has snap joints to assembly. We first introduce a control
strategy for snap assembly. Then we discuss a method for
identifying assembly states during snap assembly.
We further discuss the physical simulation method for assembly.
Lastly, we introduce an assembly planning method taking the
elasticity of parts into consideration.
Bios: Kensuke Harada received a Ph.D. from the Graduate School of Mechanical Engineering at Kyoto University in 1997. He was a research associate at Hiroshima University from 1997 to 2002. He joined the Humanoid research group of AIST in 2002. For one year from 2005 to 2006, he was a visiting scholar at the Computer Science Department of Stanford University. Currently, he is a leader of vision and manipulation research group, AIST.
Juan Rojas received his BS, MS, and PhD in Electrical and Computer Engineering from Vanderbilt University in 2002, 2004, and 2009 respectively. From 2009–2011, he served as a Visiting Scholar at Sun Yat-Sen University in China. After conducting research as a Post-doctoral Associate at National Institute of Advanced Science and Technology (AIST), Japan on 2011-2012, Dr. Rojas is appointed as an Assistant Professor at Sun Yat-Sen University since 2012. His research interests include long-term and short-term automation, cooperation and coordination of robot teams and human-robot interaction.
Eiichi Yoshida received a Ph.D degree from the Graduate School of Engineering at the University of Tokyo in 1996. In 1996 he joined the former Mechanical Engineering Laboratory, later reorganized as the National Institute of Advanced Industrial Science and Technology (AIST) in Tsukuba, Japan. He served as Co-Director of the CNRS-AIST Joint French-Japanese Robotics Laboratory (JRL) at LAAS-CNRS in Toulouse, France, from 2004 to 2008. Since 2009, he has been Co-Director of CNRS-AIST JRL (Joint Robotics Laboratory), UMI3218/CRT, Tsukuba, Japan. His research interests include robot task and motion planning, modular robotic systems, and humanoid robots.
Jean-Paul Laumond (LAAS-CNRS, France)
Title: Motion Planning for Humanoid Robots: Feedback from HRP2-14 Platform
Bio: Jean-Paul Laumond, IEEE Fellow, is Directeur de Recherche at LAAS-CNRS in Toulouse, France. He joined CNRS in 1985. He has been a co-director of the French-Japanese lab JRL from 2005 to 2008. In 2001 and 2002 he created and managed Kineo CAM, a spin-off company from LAAS-CNRS devoted to develop and market motion planning technology. In 2006, he launched the research team Gepetto dedicated to Human Motion studies along three perspectives: artificial motion for humanoid robots, virtual motion for digital actors and mannequins, and natural motions of human beings. He has been the 2011-2012 recipient of the Chaire Innovation technologique Liliane Bettencourt at Collège de France in Paris. His current project Actanthrope (ERC-ADG 340050) is devoted to the computational foundations of anthropomorphic action.
Dinesh Manocha (UNC-Chapel Hill Department of Computer Science)
Title: Motion Planning and Proximity Computations for Industrial Robots: Challenges and Lessons Learned
Yukiyasu Domae & Akio Noda (Mitsubishi Electric Corporation)
Title: Robot Programming for Assembly Tasks
Abstract: Sooner or later in the future, population in production decline will proceed all over the world. Thus, the spread of automation is a promising solution. Among them, the programming of assembly tasks is a difficult problem. A research group of the authors are working to accelerate the programming of manufacturing robots. This time, the author provides topics with the developed technologies such as an action planning method of the robotic bulk parts alignment, which is one of the classic challenges for robotics.
Bios: Yukiyasu Domae is a researcher at Advanced R&D Center of Mitsubishi Electric Corporation in Japan. He joined the current company in 2008. He received his BS in Engineering, MS and Ph.D. in Informatics from Hokkaido University in 2004, 2006 and 2012 respectively. His research interests are in robot vision, grasping and manipulation for industrial robots.
Akio NODA received the BS and MS degrees in mechanical Engineering in 1985 and 1987 respectively from Osaka University, Osaka, Japan. At 1987, he joined R&D section in Mitsubishi Electric Corporation (Hyogo, Japan). He is currently a head researcher at Advanced Technology R&D Center of Mitsubishi Electric Corporation (Hyogo, Japan). His research interests include intelligence machines, robotics, and manufacturing systems. He is a member of the Robotics Society of Japan, the Institute of Systems, Control and Information Engineers (Japan), the Society of Instruments and Control Engineers (Japan), the Japan Society of Mechanical Engineers.
Quang-Cuong Pham (Nanyang Technological University, Singapore) & Huan Liu (Mujin Inc.)
Title: Integrating Dynamics into Industrial Motion Planning
Abstract: Motion planning under geometric constraints (such as joint limits, obstacle avoidance,...) can be considered as a mature topic, as illustrated by the success of companies specialized in motion planning which have spun off from academia, such as Mujin Inc. (Japan) or Kineo (part of Siemens, Germany). By contrast, kinodynamic constraints (such as torque limits, friction limits, dynamic balance,...) still represent a major challenge, and to our knowledge, no existing commercial software to date can take into account kinodynamic constraints in a general and effective way. In this talk, we first discuss the theoretical and practical difficulties associated with kinodynamic motion planning. We then present our approach to tackle this problem, which is based on the idea of Time-Optimal Path Parameterization (TOPP). Along the way, we also evoke OpenRAVE, which is the software platform on which Mujin's applications are developed, and sketch our ongoing integration of TOPP with OpenRAVE in order to provide kinodynamic motion planning for a large range of industrial robots.
Bio: Pham Quang Cuong is currently an Assistant Professor in the School of Mechanical and Aerospace Engineering, NTU, Singapore and a Scientific Advisor to Mujin Inc., Japan. He graduated from the Departments of Computer Science and of Cognitive Sciences of École Normale Supérieure rue d'Ulm, France, in 2007, and obtained his PhD in Neuroscience from Université Paris VI and Collège de France in 2009. From 2010 to 2013, he was a JSPS Postdoctoral Fellow at the University of Tokyo.