Research: Network Systems
Connectivity Perserving Networks

Many mission scenarios can benefit from networks of autonomous agents functioning with collective goals that require collaborative actions. Often these networks are formulated in a decentralized manner to provide robustness against the single point of failure of centralized computing and decision making. Decentralized networks rely heavily on communications with peers for feedback regarding progress towards the collective objective. NCR efforts focus on decentralized network control where network connectivity is considered as a constraint. Efforts also consider the further challenge when communication is intermittent or event triggered, as a means to reduce communication bandwidth.


Event-triggered Multi-agent Systems

Multi-agent systems (MASs) are traditionally designed under the assumption that state feedback is continuously available with each agent continuously communicating with its neighbors. Switched/hybrid systems methods, can be used to obtain performance certificates, scalability bounds, and timing conditions for systems with discontinuities arising from communication channel dynamics, vehicle dynamics, and decision logic. Using these tools incorporated with Lyapunov analysis methods, we investigate event- and self-triggered control of MAS where intermittent communication dictated by event triggers and timing conditions developed from switched/hybrid systems methods and Lyapunov-based analysis. Such triggered communication eliminates the assumption of continuous communication with implications of reduced communication bandwidth and energy resources and improved flexibility for operations in contested environments.


Herding

For various applications (e.g. adversarial target tracking and crowd control), there may be agents in the system that are not directly controllable, posing a unique challenge to accomplish certain objectives. In such example scenarios, directly controllable agents are tasked with interacting with other target agents in the system in an effort to regulate them to a goal location. One approach is to model interactions between the herder, i.e., the controllable agent, and the target agents for which a control law cannot be directly implemented, and design a control law such that the regulation objective is achieved.


Jamming

Wireless communication networks are widely used in commercial, industrial, and military applications. Due to the openness of the wireless medium, such networks are vulnerable to interference, failure, and attack. Various jamming techniques have been developed to disrupt different layers of the protocol stack in wireless networks, such as causing errors in the reception of data in the physical layer, blocking transmission of data at the MAC layer, or taking advantage of the periodicity of many routing protocols and performing jamming at the network layer. NCR efforts focus on determining optimal locations for network jamming, and exploiting mobile jamming as a means to reduce the number of overall jamming agents.


Power Control

Power control in a code-division multiple access (CDMA) based cellular network is a challenging problem because the communication channels change rapidly because of multipath fading. These rapid fluctuations cause detrimental effects on the control efforts required to regulate the signal-to-interference plus noise ratios (SINRs) to the desired level. NCR efforts in this area focus on the development of power-control algorithms that can adapt to rapid changes in the channel gain caused by multipath fading. Key contributions include the development a controller for the reverse link of a CDMA cellular system, which includes the effects of fast fading, and use a Lyapunov-based analysis to show convergence of the SINR error. A linear prediction filter that utilizes local SINR measurements and estimates of the Doppler frequency, derived from local SINR measurements, is used to improve the estimate of the channel fading used in the controller.


Social Networks

Social interactions influence our thoughts, opinions and actions. Technological advances in social media provide more rapid, convenient, and widespread communication among individuals, which leads to a more dynamic interaction and influence. NCR research efforts focus on the study of social interactions within a group of individuals composed of influential social leaders and followers. Each person is assumed to maintain a social state, which can be an emotional state or an opinion. Followers update their social states based on the states of local neighbors, while social leaders maintain a constant desired state. Social interactions are modeled as a general directed graph where each directed edge represents an influence from one person to another. Motivated by evidence in social science literature that people respond to events based on a moral memory, efforts also focus on the development and use of networked fractional order models for human behavior.




Sponsors
Ongoing Projects
AFOSR: Center of Excellence in Assured Autonomy in Contested Environments
AFOSR: A Switched Systems Approach for Navigation and Control with Intermittent Feedback
AFRL: Privileged Sensing Framework

Completed Projects
NSF NeTS: Small: Network Connectivity and Security for Cooperative Autonomous Vehicles
AFRL: Mathematical Modeling and Optimization Institute

Related NCR Network Systems Publications