Kaustubh Sinkar

Network science based approaches to design and analyze MANETs for military applications

Mobile ad hoc networks have become the basis of the militarys network-centric warfare (NCW) approach. However, for NCW to be successful, it is imperative that the networks be designed in a robust manner with the capability to produce consistent predictable results despite the uncertainties of the underlying environment. This underscores the need for formal systematic methodologies to design and predict performance of such networks. The challenges of mobile ad hoc networking combined with those associated with the stringent requirements posed by NCW systems, however, are daunting, and thus no systematic design techniques for NCW system design exist. To address this problem, a joint project was initiated between CERDEC and Telcordia Technologies to develop the Network Engineering Design Analytic Toolset (NEDAT) - a toolset that applies network-science-based approaches to design MANETs for use in NCW. Rooted in formal/analytic techniques, NEDAT can be used to design MANETs for use in NCW given information about available resources and performance objectives, analyze performance of a given NCW network, and understand design trades.

Cognitive tactical network models

Unlike commercial MANET applications, tactical networks are typically hierarchical and involve heterogeneous types of radio communications. Future tactical networks also require cognitive functions across the protocol stack to exploit scarce spectrum and dynamically adapt functions and configuration settings. In this work we highlight the need for novel design tools for cognitive tactical networks. We define a system design model that will provide the foundation for generic network design problem formulations via the use of cognitive techniques covering both dynamic frequency adaptations and machinelearning- related aspects of cognition. We use the system model to identify several potential cognitive design knobs and describe how the different design knobs can potentially be adjusted at different timescales of operation. These knobs are used in formulating a cognitive network design problem. Finally, we discuss how a network designer can potentially benefit from the proposed model result, a cognitive network design toolset we have recently developed.

Resource Allocation and Performance Study for LTE Networks Integrated with Femtocells

Long-Term Evolution (LTE) networks comprising conventional cellular macrocells plus user-installed femtocells offer an economically viable solution to achieving high user capacity and upgrading to future fourth-generation systems. With the growing impetus for frequency reuse, the capacity of each user depends on not only the power spectral density of its own, but also on those of others in neighboring cells. Mitigating interference among macrocells and femtocells requires allocating physical resource dynamically in response to channel conditions. In this paper, we formulate the resource allocation problem as a utility optimization and develop a distributed algorithm for joint power control and user scheduling. The algorithm makes novel use of a class of fairness measures for determining user scheduling and is shown to be very efficient for realistic network parameters. Additionally, using a practical model for the LTE air interface that captures geographic distribution of users and buildings, we provide for a framework that allows comparison of different resource allocation algorithms. A variety of problem formulations, including femtocell density, resource tradeoff, and complexity-optimality tradeoff are derived and analyzed using a geometry-based stochastic LTE air interface model. Our analysis also offers useful guidelines for the planning and design of macrocells and femtocells.

C-NEDAT: A cognitive network engineering design analytic toolset for MANETs

Future force networks of the types envisioned for the network centric warfare (NCW) paradigm will be highly diverse, with the diversity spanning a wide range of (a) requirements (e.g., need for capacity, connectivity, survivability), (b) resources (e.g., radios with widely different capabilities and ‘smart’ (e.g., Software Defined Radios (SDRs)), and (c) environments (e.g., urban, rural). The need to facilitate robust and adaptable communications in such networks has in turn triggered research in the area of cognitive networks that have the ability to ‘learn’ and generate real-time control actions to adapt to the wide diversity of requirements, resources and environments. However, the combination of diversity and “smart” networking exacerbates the problem of generating reliable and robust network designs. We present in this paper, our work on the use of cognitive mechanisms to assist with the design and analysis of robust NCW-like networks. Based on formal network-science based approaches, our Cognitive Network Engineering Design Analytic Toolset (C-NEDAT) provides for a systematic way to design, analyze and maintain robustness of future force MANETs. We provide in this paper an overview of the key functional modules and design capabilities of C-NEDAT and present example results.

Tuning of reinforcement learning parameters applied to OLSR using a cognitive network design tool

In wireless mesh networks, with the standard Optimized Link State Routing (OLSR) metric (i.e. hop count), traffic is routed on the shortest path without considering factors such as traffic distribution and link capacities. Consequently, some nodes may get overloaded from the uneven utilization of network resources. OLSR can be modified to use other link cost metrics, with route selection based on lowest cost path. With delay as the metric, OLSR reduces average round trip time but the load-aware routes may cause wide variance in delay and packet reordering due to route oscillations. We describe a new hybrid routing approach that combines the strength of a) link state routing (e.g. fast convergence), b) load-aware routing (e.g., avoiding congested paths) and c) cognitive routing (e.g. learning to avoid path oscillations). In particular, we investigate the use of Q-learning with OLSR to increase network capacity and reduce congestion delay. We present simulation results for a 36 node dynamic mobile ad hoc network, with standard OLSR, a non-cognitive load-aware OLSR (OLSR-D) and our new hybrid cognitive load-aware OLSR (OLSR-Q). We show that OLSR-Q >; OLSR-D >; OLSR in terms of reducing delay and increasing network capacity. Furthermore, we show that, unlike conventional cognitive Q-routing protocols, our hybrid approach does not reduce performance at low load. Although OLSR-Q can significantly reduce delay and improve capacity, the learning time can reduce connectivity and the distribution of more link state information can reduce raw link capacity. We show how adding OLSR, OLSR-D and OLSR-Q as routing options into the Cognitive Network Engineering Design Analytic Toolset (C-NEDAT), we can select the best routing protocol and parameters (e.g., learning rate) for a given network and its mission. We verify simulation performance improvements by implementing the OLSR-Q in on a 9 node wireless testbed.

Scalable Registration and Discovery of Devices in Low-Bandwidth Tactical Networks

Management of RF Network and Tasking Infrastructure (MARTI) is a distributed system that discovers, tracks, coordinates, and manages the reception and transmission capabilities of an overlay network of RF devices. To enable scalable discovery and tracking, we employ a dynamic hierarchical domain management technique with features targeted for low-bandwidth networks. The dynamic hierarchy achieves the efficiency benefit of hierarchical systems, while avoiding single points of failure by dynamic reelection of domain manager (DM) nodes in response to node failures and disconnections. The Domain Announcement Protocol (DAP) employed for this purpose autonomously creates and organizes domains in a dynamic structure, supports splits and merges of domains, and employs a rich set of metrics for DM election, which enables DAP to adapt to loss of links or node failures. Furthermore the criteria used for the dynamic election of DMs are tuned so that DMs can also serve as distributed and survivable Directory Agents (DAs). Thus the DAP DMs which support the formation of domains are also used to realize the MARTI directory agent functions, which is an unusual example of efficient cross-layer functional re-use.

Automated design algorithms for tactical wireless networks

To guide users who attempt to deploy wireless networks in military applications, there is an evolving need for developing systematic methodologies to analyze/predict the performance of mobile ad hoc networks (MANETs). In addition, the advance in cognitive networking research provides opportunities for exploiting unused spectrum to optimize throughput of MANETs. However, with the increasing number of parameters/constraints, there is even a more demanding need to develop automated methodologies to design/tune such networks. In this work, we study the concepts and challenges for automatic design/re-configuration of cognitive MANETs, in addition to proposing design automation algorithms. The paper is divided into two parts. In the first part, we describe the design objectives, imposed constraints, and involved parameters in MANET design. We discuss how cognitive techniques can be employed to exploit the unused spectrum in military architectures. We then discuss the challenges that face the design/re-configuration of a cognitive network and their implications at different network layers. We also describe possible implementation options for designing MANETs that employ cognitive features at all layers. In the second part of this work, we propose design automation algorithms for optimally setting parameters to achieve a desired objective and satisfy certain constraints. Despite providing the optimal configuration, the simple approach of testing all possible combinations of parameter settings has significant time complexity (the COMB approach). Thus, we propose a novel heuristic (Sequential Parameter Optimization or SEPO) for searching through the possible parameter settings and selecting the best design options. SEPO is efficient in terms of both convergence speed and parameter tuning. We also discuss the foundation for using supervised learning to speed up the design (search) process. By evaluating realistic design of military-like scenarios that require optimizing a diverse set of metrics, we show that SEPO generates comparable results to the optimal, straightforward (slow-converging) COMB approach that is based on exhaustive search.