Cooperative Control of Networked Systems

Multi-agent systems have recently emerged as an inexpensive and robust way of addressing a wide variety of tasks ranging from exploration, surveillance and reconnaissance, to cooperative construction and manipulation. The agent here refers to the autonomous ground, aerial or aquatic robots which require a significant amount of information gathering, data processing, and decision making when performing tasks, without explicit human control. The success of these stories relies on efficient information exchange and coordination between the members of the team. 

My main interest is to develop decentralized controllers for a group of autonomous agents to perform coordinated global tasks using local information. Two key technical challenges are:

  • Network Connectivity: To achieve a cooperative objective a multi-agent system typically requires that the agents collaborate over a communication network. However, wireless networks are severely constrained in terms of bandwidth and range. In view of this, assuming limited communication and sensing capabilities on agents, each agents knows the positions of only those nodes within its sensing range and can only communicate with nodes within its communication range. Thus, the objective must be accomplished while ensuring that specified nodes stay within each other’s sensing and communication ranges and that the overall communication network stays connected.
  • Control Design: Decentralized control is an approach in which each agent makes an independent decision based on local information from neighboring neighbors. However, di.culties arising from performing required mission objectives for the global network using local feedback can cause the network to partition. When the network partitions, communication between groups of agents can be permanently severed leading to mission failure.

 

Network Connectivity & Decentralized Formation Control

A decentralized control method is developed to enable a group of agents to achieve a desired global configuration while maintaining global network connectivity and avoiding obstacles, using only local feedback and no radio communication between the agents for navigation. To navigate the agents to a desired configuration while avoiding obstacles, the decentralized controller is developed based on the navigation function formalism. By proving that the proposed controller is a qualified navigation function, convergence to the desired formation is guaranteed.

  • Global asymptotic convergence to the desired configuration using only local sensing information
  • The developed method ensures the desired communication links remain connected for all time (Wireless communication is available among agents for other tasks.)
  • No radio communication between the agents for navigation is required, which enables stealth modes of operation
  • Collision avoidance with other agents and static obstacles
  • Assume a connected initial graph with desired neighborhood between agents

 

Related Publication: 

Z. Kan, A. Dani, J. M. Shea, and W. E. Dixon, Network Connectivity Preserving Formation Stabilization and Obstacle Avoidance via A Decentralized Controller,” IEEE Transactions on Automatic Control, Vol 57, No. 7, pp. 1827-1832 (2012). [pdf] [simulation video]

Z. Kan, A. Dani, J. M. Shea, and W. E. Dixon, “Ensuring Network Connectivity During Formation Control Using A Decentralized Navigation Function,” Military Communications Conference (MILCOM), San Jose, CA, 2010, pp. 954-959. [pdf]

 

Motivated to steer a group of agents to a desired configuration from any given initially connected graph, A navigation function based decentralized controllers are then developed to maintain the underlying network connected, and guarantee the convergence of the system to the desired configuration, as well as collision avoidance with other agents and moving obstacles. An information flow is proposed to specify the necessary communication among agents. Based on the approach of information flow, each agent is able to choose a short path to reach the desired agent in the information graph by dynamically building new communication links or breaking existing links.

 

  • Global asymptotic convergence to the desired configuration using local information
  • Any given initially connected graph
  • Maintenance of the network connectivity all the time
  • Collision avoidance with other agents and moving obstacles

 

Related Publication: 

Z. Kan, A. Dani, J. M. Shea, and W. E. Dixon, “Information Flow Based Connectivity Maintenance of A Multi-agent System During Formation Control,” IEEE Conference on Decision and Control, Orlando, FL, 2011,  pp. 2375-2380. [pdf]

 

Network Connectivity & Rendezvous

To achieve a cooperative objective a multi-robot system typically requires that the robots collaborate over a communication network. A control strategy is designed for repositioning and reorienting a group of wheeled robots with nonholonomic constraints and limited communication and sensing capabilities. Each robot knows the positions of only those nodes within its sensing range and can only communicate with nodes within its communication range. Thus, the objective must be accomplished while ensuring that specified nodes stay within each other’s sensing and communication ranges and that the overall communication network stays connected. To achieve these objectives, we develop a dipolar navigation function and corresponding time-varying continuous controller. We show that if the network is initially connected, the controller maintains the specified communication links at all times while moving the robots into the specified positions and orientations. We consider the particular application of moving the robots to a common rendezvous point with a specified orientation.

Related Publication:

Z. Kan, A. Dani, J. M. Shea, and W. E. Dixon, Ensuring Network Connectivity for Nonholonomic robots During Rendezvous,” IEEE Conference on Decision and Control, Orlando, FL, 2011, pp. 2369-2374. [pdf]

Z. Kan, Justin Klotz, Teng-Hu Cheng, and W. E. Dixon, Ensuring Network Connectivity for Nonholonomic Robots During Decentralized Rendezvous,” American Control Conference, Montréal, Canada, 2012, pp. 3718-3723. [pdf]

 

Network Connectivity & Flocking with Switching Topology

 

A two level control framework is developed for connectivity maintenance and cooperation of multi-agent systems. Each agent is equipped with an omnidirectional camera and wireless communication capabilities. Image feedback is the primary method to maintain connectivity among agents with wireless communication that is only used to broadcast information when a specific topology change occurs. All agents in the team are categorized as clusterheads or regular nodes. A high level graph is composed of all clusterheads and the motion of the clusterheads is controlled to maintain existing connections among them. A low level graph composed of all regular nodes is controlled to maintain connectivity with respect its specific clusterhead.

 

Related Publication:

Z. Kan, A. Dani, J. M. Shea, and W. E. Dixon, Ensuring Network Connectivity for Nonholonomic robots During Rendezvous,” IEEE Conference on Decision and Control, Orlando, FL, 2011, pp. 2369-2374. [pdf]

Z. Kan, Justin Klotz, Teng-Hu Cheng, and W. E. Dixon, Ensuring Network Connectivity for Nonholonomic Robots During Decentralized Rendezvous,” American Control Conference, Montréal, Canada, 2012, pp. 3718-3723. [pdf]

Z. Kan, S. Subramanian, J. M. Shea, and W. E. Dixon, Vision Based Connectivity Maintenance of a Network with Switching Topology,” IEEE International Symposium on Intelligent Control part of the Multi-Conference on Systems and Control, Yokohama, Japan, September 2010, pp.1493-1498. [pdf]