The future of multiple-robot research and its multiple identities


Workshop at 2014 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

September 18th, Chigago, IL, USA




Invited talks


Rachid Alami
LAAS-CNRS, France

Title:
Decisional issues for multi-robot systems: Towards teams of humans and robots

Abstract:
I will discuss decisional issues that are required in order to equip a robot with the capabilities to behave autonomously as a  teammate in a team composed of humans and robots.
This decisional capabilities cover issues ranging from mission planning  to effective conflict free and coordinated execution in a dynamic environment. We envisage a combination of local individual planning and coordinated decision for the robot to decide what it has to do and for incremental plan adaptation to the multi-robot context and also to the human context.
The assumption is that, in a complex "system" composed of humans and of several autonomous robots equipped with their own sensors and effectors, the ability of a given robot, to achieve a given task in a given situation can be best computed using a planner. Indeed, we claim that the robots must be able to plan/refine their respective tasks, taking into account the other agents (humans and robots) intentions and plans as planning/refinement constraints, and thus producing plans containing coordinated and cooperative actions.

Program





Magnus Egerstedt
Georgia Institute of Technology, USA

Title:
Eulerian Swarms

Abstract:
The traditional approach to controlling multi-robot systems is to view the individual robots as agents, whose dynamics are to be controlled; autonomously or through human inputs. This can be thought of as following the Lagrangian approach to fluid-dynamics, where the movements of individual particles are characterized. In this talk, we contrast this with the “Eulerian” approach, whereby individual robots do not matter. Instead, what is characterized are overall robot “flows” or “densities”, which gives human operators the ability to control the multi-robot system without having to worry about the motions of individual robots.

Program





Randy Freeman
Northwestern University, USA

Title:
Dynamic average consensus for multi-robot coordination

Abstract:

There are many robot coordination algorithms which can benefit from dynamic average consensus, that is, the ability for robots to compute accurate running estimates of the current global average of a collection of local, time-varying input signals. Examples include Kalman filtering for environmental estimation and motion control for maintaining communication connectivity among a group of mobile robots.  To be useful for robot applications, the computation of such global averages should be distributed, scalable, and robust. In this talk we will review some recently proposed dynamic average consensus estimators, show how they are related  o each other, and discuss how the classical Bode integral constraints lead to limits on their achievable performance.

Program





Vijay Kumar
University of Pennsylvania, USA

Title:
Coordination, Cooperation, and Collaboration in Multi Robot Systems

Abstract:
tbd

Program





Jonathan P. How
Massachusetts Institute of Technology, USA

Title:
Decentralized Autonomy: Planning and Learning

Abstract:
The talk will present algorithmic solutions to some key challenges in multiagent autonomous learning and planning.  Reliable robust planning depends on the availability of models that accurately represent the environment and its evolution. One approach for efficiently learning probabilistic models in dynamic environments utilizes Bayesian nonparametric models that provide the flexibility to learn both the structure and the parameters of a model, which are often very difficult to determine a priori. For example, in motion planning domains, Gaussian processes (GPs) are used to represent the trajectory velocity fields of obstacles (static & dynamic) in the environment, a Dirichlet process GP mixture (DP-GP) is used to learn the number of motion models and their velocity fields, and the dependent Dirichlet process GP mixture (DDP-GP) augments these models by capturing the obstacles' temporal evolution. In practice, these techniques have been used to learn models of motion/intent behaviors of automobiles and pedestrians to improve the performance of autonomous vehicles.
Multiagent scenarios introduce other challenges when limited communication, due to hardware constraints or features of the operation environment, do not allow for centralized learning and planning. Learning in these environments forces individual agents to construct and share compact representations of data rich observations so that manageable communication levels can be achieved while allowing for model consistency across agents. Likewise for planning, agents need to be explicitly aware of, and account for, the effect of limits in inter-agent bandwidth on planning. This talk will present algorithms for both aspects of decentralized multiagent operations.

Program





Ani Hsieh
Drexel University, USA

Title:
Going with the Flow: A Case for Multi-Robot Systems and Distributed Mobile Sensing

Abstract:
The last few years have seen an explosion in the interest of micro autonomous vehicles.  These platforms must operate with significant size, weight, and power constraints.  However, as vehicle sizes become smaller, vehicle dynamics become more tightly coupled to the surrounding environmental dynamics and making vehicle control hard.  These are challenges continually faced by underwater robots due to the high inertia environment they operate in.  In this talk, I will present our work in leveraging the spatio-temporal sampling capabilities of a team of robots to track and estimate various physical processes that give us insight into the dynamics of a geophysical fluid environment like the ocean.  The ability to estimate and predict the environmental dynamics can significantly increase the maneuverability and energy-efficiency of single micro autonomous vehicles since vehicles can leverage the environmental forces to navigate in the workspace.  Furthermore, by understanding the dynamics of the environment, autonomous robots can continuously adapt to changing environmental conditions as they execute their assigned tasks.   The tracking and estimation of any geophysical fluid dynamics is of course a challenging problem, since geophysical flows like the ocean spans large physical scales, is stochastic, and time-varying.   Nevertheless, this is a problem is one that can only be formulated in the multi-robot context. 

Program





Volkan Isler
University of Minnesota, USA

Title:
Robotic Data Collection in Environmental and Agricultural Applications

Abstract:
Robots are tremendously effective in controlled, factory-like environments. In contrast, making robots operate in dynamic and complex environments remains a major challenge.  Robotic Sensor Networks composed of robots and wireless sensing devices hold the potential to revolutionize environmental and agricultural sciences by enabling data collection across expansive environments, over long, sustained periods of time. In this talk, I will report our progress on building such systems for two applications. The first application is on monitoring invasive fish (common carp) in inland lakes. In the second application, an unmanned aerial vehicle and a ground vehicle act as data mules and collect data for precision agriculture.  After presenting results from field experiments, I will focus on the problem of designing robot trajectories to collect data from possibly mobile targets, and present recent results.

Program





Anibal Ollero
University of Seville, Spain

Title:
Multiple cooperative aerial robots: From intentional cooperation to the control of physical interactions

Abstract:
The presentation will start with a classification of multiple aerial robot architectures and a general model for cooperation.
Then, particular methods for the cooperation of aerial robots developed in several European projects coordinated by the author will be presented.
The presentation will include decentralized solutions for cooperative patrolling strategies of heterogeneous aerial robots dealing with communication constraints, dynamic task allocation for surveillance, and tracking methods taking into account uncertainties.
Finally, cooperation involving physical interactions will be analysed. In this case the cooperation for transportation and manipulation with aerial robots will be considered.
The presentation will include simulations, experiments in the FADA-CATEC indoor testbed, and outdoor experiments.

Program





Lynne Parker
University of Tennessee, USA

Title:
Characterizing Commonalities in Multi-Robot Systems Research

Abstract:
Multi-robot systems have been a topic of study for more than two decades.  Many important advances have been made in the field, but less attention has been paid to understanding an overriding structure to the technical approaches used to develop working multi-robot systems.   Given the multitude of research activities worldwide that involve multiple robots, it is interesting to identify fundamental questions and technical approaches that unify these various studies, other than the fact that they employ more than one robot.   This talk will suggest some over-arching paradigms to the software control of these systems, such as the bioinspired, organizational, and knowledge-based paradigms.  Similarities and differences in such paradigms will be discussed, followed by a discussion of how the choice of paradigm can impact the solution strategy for given technical challenges in multi-robot systems.  The talk will conclude with a challenge to the community to expand our understanding of overriding paradigms and compile the many existing multi-robot point solutions into a "guidebook" that facilitates the movement of multi-robot systems from theory to practice.

Program





Mac Schwager
Boston University, USA

Title:
Characterizing Commonalities in Multi-Robot Systems Research

Abstract:
Multi-robot systems have been a topic of study for more than two decades.  Many important advances have been made in the field, but less attention has been paid to understanding an overriding structure to the technical approaches used to develop working multi-robot systems.   Given the multitude of research activities worldwide that involve multiple robots, it is interesting to identify fundamental questions and technical approaches that unify these various studies, other than the fact that they employ more than one robot.   This talk will suggest some over-arching paradigms to the software control of these systems, such as the bioinspired, organizational, and knowledge-based paradigms.  Similarities and differences in such paradigms will be discussed, followed by a discussion of how the choice of paradigm can impact the solution strategy for given technical challenges in multi-robot systems.  The talk will conclude with a challenge to the community to expand our understanding of overriding paradigms and compile the many existing multi-robot point solutions into a "guidebook" that facilitates the movement of multi-robot systems from theory to practice.

Program





Kosuke Sekiyama
Nagoya University, Japan

Title:
Cognitive issues of vision-based multi-robot cooperation

Abstract:
Most work of vision-based multi-robot system has focused on the architecture for cooperation and controlling the behaviors. That is surely important, however, the argument on cognitive aspects are very limited. Visual information is highly dependent on the robot-friendly environment using artificial markers, where the observation target is predefined, and in many cases, the objects leading to misrecognition are eliminated from the scene.
There are unique cognitive issues peculiar to multi-robot systems. Each robot for cooperative task is often required to share the target of object, which is not predefined. Each robot is located at different position and has a different viewpoint. Hence, it has a different perspective for the target and surroundings. From the different perspective, the appearance of the same object would be seen differently, because of some lighting conditions, partial or complete occlusion, and a visual deformation of the shape. Therefore the effective feature for a robot to identify the object is not necessarily effective for another robot to be shared.
In this talk, the fundamental cognitive issues in the multi-robot system are clarified as following three points; that is (1) cognitive sharing of target, (2) ROI(region of interest) selection, and (3) viewpoint selection. Several main results to cope with these issues are presented including the consensus making algorithms by selection of viewpoint-invariant representations, and the viewpoint selection scheme for a monitoring robot.

Program