2014 IEEE ICRA
Half-day Workshop
2014 IEEE ICRA
Half-day Workshop
Afternoon Session, Sunday, June 1st, 2014, Room: S426
Hong Kong, China
On the Centrality of Decentralization in Multi-robot Systems:
Holy Grail or False Idol?
Half-day ICRA Workshop
Hong Kong Convention and Exhibition Center, Hong Kong, China
Sunday, June 1st, 2014, 2pm—6pm
Room: S426
Motivations and Objectives
From multi-robot decentralized theory to practice: after more than a decade of theoretical studies and advancements on decentralized estimation and control for multi-robot systems and distributed algorithms for networks of robotic agents, several real-world experiments have been carried out and the gap between theory and practice is becoming small.
But is decentralization a real benefit for teams of multi-robots, or is it more a pure theoretical challenge? What of the various decentralized features have found a practical relevance? In robotics and distributed control, decentralization is typically achieved at the cost of local optimality and slow convergence rates, while Nature shows several examples of efficient and highly dynamic behaviors of large scale animal populations. Is this evident gap due to a methodological and insufficiency of our theoretical tools or to a technological lack, i.e., in terms of perceptual and actuation performances?
This workshop will host some of the most renowned researchers in the multi-robot field and will let them both present their most recent results and thoroughly discuss the aforementioned questions and their fundamental implications in future multi-robot research.
Intended Audience
Academic researchers in the robotics and automation communities who are currently active (or intend to become) in the field of multi-robot planning, estimation and control. One of the objective of the Workshop is also to bring together researchers in academia, government, and industry.
List of topics
This workshop is supported by the IEEE RAS Networked Robots Technical Committee.
The workshop topics cover, but are not limited to:
- multi-robot/multi-agent systems
- distributed algorithms
- swarm robotics
- mobile sensor networks
- mobile communication/ad-hoc networks
- heterogeneous multi-robot platforms
- decentralization vs. centralization
- scalability vs. performance and reliability
- decentralization vs. efficiency
Schedule
download the schedule of the two sibling workshops as pdf
14.00 - 14.10 Welcome
14.10 - 15.30 Keynote Session 1:
14:10-14:50 Alcherio Martinoli (EPFL, Switzerland)
Modeling and Control of Distributed Stochastic Robotic Systems
Abstract: Technological advances in communication, embedded computing, energy storage, sensors and actuators enable an increasingly ubiquitous deployment of distributed, mobile, robotic systems in the real world. However, often such systems are severely constrained in their resources by cost, volume, or mass considerations imposed by the targeted application. Some of these constraints might even enforce a centralized or a decentralized control solution but they typically result in an increased stochasticity of the node behavior that has to be captured and controlled with appropriate methods in order to obtain a more predictable behavior at the collective system level. In this seminar, I will describe a few recipes that allowed us to achieve such result under peculiar scenarios. In particular, I will focus on a multi-level modeling framework that has been instrumental for efficiently applying a number of control design and optimization techniques. I will support the discussion with a few specific case studies concerned with spatial aggregation and assembly of nodes to illustrate such methods. Despite the experimental scenarios related to these case studies are characterized by different environmental templates and capabilities of the individual nodes in terms of computation, mobility, sensing, and actuation, I will show that the overall multi-level modeling framework remains the same. Finally, I will conclude my seminar with some of the lessons we learned over the last eighteen years of research in this area and extrapolate some hints for future research directions to overcome limitations of the current modeling and control methods.
14:50-15:30 Kostas Bekris (Rutgers University, USA)
Properties of Planning Methods for Multi-Robot Systems
Abstract: This talk will review a variety of approaches to motion planning for multi-robot systems, ranging from centralized to decentralized ones. The focus is on the type of guarantees that can be argued in each case, relating to properties such as safety, completeness, optimality, deadlock/livelock avoidance and information/communication requirements. Frequently many of these guarantees are sacrificed, when a fully decentralized, online method is preferred. But even in the case of centralized solution, such guarantees may be difficult to achieve in a computationally efficient manner given the complexity of the underlying challenge.
15.30 - 16.00 Coffee Break & Interactive Session
R. Grieder, J. Alonso-Mora, C. Bloechlinger, R. Siegwart and P. Beardsley (ETH Zurich, Disney Research center Zurich)
Multi-robot Control and Interaction with a Hand-held Tablet
S. Kim, S. J. Guy, W. Liu, D. Wilkie, R. W. H. Lau, M. C. Lin and D. Manocha (Univ. of North Carolina, Univ. of Minnesota, City University of Hong Kong)
Predicting Pedestrian Trajectories for Robot Navigation
D. Sofge, M. Kuhlman, N. Sydney and D. Paley (Naval Research Laboratory, UMD)
Mobile Autonomous Navy teams for Information Surveillance and Search (MANTISS)
V. Digani, L. Sabattini, C. Secchi and C. Fantuzzi (Univ. of Modena and Reggio Emilia)
Decentralized coordination enhanced by centralized information: multiple AGVs in industrial application
R. K. Williams, A. Gasparri, and G. S. Sukhatme (Univ. Southern California, Univ. Roma 3)
Rigid Networks for Feasible Collaboration and a Taxonomy of Interconnected Systems
G. Gioioso, A. Franchi, G. Salvietti, S. Scheggi and D. Prattichizzo (University of Siena, Italy; IIT, Italy; LAAS-CNRS)
A Tele-operated Swarm of UAVs for Cooperative Grasping and Manipulation
E. Castello, T. Yamamoto, Y. Nakamura and H. Ishiguro (Osaka University; CiNet)
Foraging in Real and Simulated environments for a Robotic Swarm based on an Adaptive Response Threshold Model
P. Stegagno, C. Massidda, and H. H. Bülthoff (MPI for Biol. Cybernetics)
Object Recognition in Swarm Systems: Preliminary Results
16:00 - 17:20 Keynote Session 2:
16:00-16:40 Kejian Wu (presenter) / Stergios Roumeliotis (University of Minnesota, USA)
Decentralized multi-robot cooperative localization under communication constraints
Abstract: We consider the problem of solving estimation problems in a decentralized way, with the focused application of cooperative localization. In the absence of bandwidth constraints, the optimal solution is given by the distributed MAP estimator. However, when the communication bandwidth is limited, we advocate optimally quantizing and processing measurements. This requires i) selecting the optimal bits to quantize available measurements so that the information communicated is maximized, and ii) determining the optimal way to process these bits. In this talk, we present our solutions to the above questions: The optimal quantization thresholds per analogue measurement, the method to optimally allocate bits among multiple measurements, the MMSE and MAP estimators for processing bits, as well as a hybrid estimation framework that incorporates both local analogue and remote quantized measurements.
16:40-17:20 Filippo Arrichiello (presenter) / Gianluca Antonelli
(University of Cassino and Southern Lazio)
Experiences of (de)centralized behavioral control for multi-robot systems
Abstract: In this talk we will present some experimental results gained with behavior-based coordination control strategies for multi-robot systems developed at the University of Cassino and Southern Lazio. In particular, we will report the experimental results of missions executed with a centralized systems composed of wheeled robots, and more recent results with ground distributed multi-robot systems. We will then present results of experiments with marine platforms, belonging to research partners, performed in the framework of joint research collaboration and projects. The talk will end with a discussion concerning some features and constrains of (de)centralized systems.
17:20 - 18:00 Panel discussion
Organizers
Antonio Franchi
antonio.franchi[a]laas.fr
Centre National de la Recherche Scientifique (CNRS)
Laboratoire d’Analyse et d’Architecture des Systèmes (LAAS)
http://homepages.laas.fr/afranchi/
Paolo Robuffo Giordano
prg[a]irisa.fr
Centre National de la Recherche Scientifique (CNRS)
Lagadic Team, IRISA
http://www.irisa.fr/lagadic/team/Paolo.Robuffo_Giordano.html
Sibling workshop
This half-day workshop is intended to be a “sibling” event of the other ICRA2014 half-day workshop “Crossing the Reality Gap: Control, Human Interaction and Cloud Technology for Multi- and Many-Robot Systems”
Endorsers
- Filippo Arrichiello and Gianluca Antonelli (University of Cassino and Southern Lazio, Italy)
- Kostas Bekris (Rutgers University, USA)
-Raffaello D’Andrea (ETH, Switzerland) (subject to availability)
-Volkan Isler (University of Minnesota, USA) (subject to availability)
- Alcherio Martinoli (EPFL, Switzerland)
- Stergios Roumeliotis (University of Minnesota, USA)
- Gaurav Sukhatme (University of Southern California, USA)
Deadlines
February 28th, 2014 Call for Contribution
March 31th, 2014 Abstract Submission Deadline
April 15th, 2014 Notification of acceptance
Author Instructions
Possible contributions:
- Extended abstract (from 2 to 6 pages) for addressing a specific algorithm/result within the workshop topics. This contribution will be allocated a poster presentation or a talk after the review process.
- Extended abstract (roughly 2 pages) for proposing a contributed talk addressing, from a more general point of view, the main workshop theme, i.e., “Decentralization in multi-robot systems”. We welcome contributions discussing, also in a provocative way, the various pros/cons of decentralized solutions, and help stimulating the discussion during the workshop.
An accompanying video is optional and can also be provided as a weblink (e.g., youtube).
All contributions must be sent as a pdf file to antonio.franchi@laas.fr and prg@irisa.fr
Presentation of Accepted Contributions
Accepted contributions will be split in two groups: contributed talks and posters.
Contributed talks will be allocated a time slot of 20 min. Posters will be presented during the afternoon coffee break.