The ActComm project has several subprojects:
D'Agents is a mobile-agent system that focuses on support for multiple languages, security, fault tolerance, performance, and the ability to operate effectively in volatile, wireless networks. It will be the middle layer of the ActComm infrastructure, sitting on top of the network services but below the planning, learning, resource discovery, and information-retrieval services.
More information on D'Agents can be found at http://www.cs.dartmouth.edu/~agent.
We propose using market-based control methods to control resource allocation for mobile agents that travel across administrative domains. Agents buy necessary resources from host sites and may sell their own services to users and other agents. We have a simple currency system supported by virtual banks, and an arbitration mechanism. Our most recent efforts have focused on incentive-compatible pricing and allocation algorithms to allocate priority based resources (e.g., CPU and network interface time).
More information on market-based control can be found at http://www.cs.dartmouth.edu/~agent/research/market/.
Investigator: Vastola, RPI
We are developing a Wireless Network Environment Sensor to detect and, under certain conditions, predict wireless network link failures. Our goals are to:
Enable network-aware applications (including Dartmouth's D'Agents platform) to determine the state of the wireless network as seen from its host and to receive network status updates in order to make better decisions about when, where, if, and how to transmit and receive information across the network.
Sense the network status in a platform-independent manner using SNMP MIB variables. If required, recommend additional variables be included in the MIB for IEEE standard 802.11.
Introduce no additional load on a heavily-used link.
Most current computer applications are insensitive to changing network conditions. With the growing demand for wireless. satellite, and other highly volatile computer communications networks, however, applications that are robust in the presence of network volatility must be designed and built. Network-robust applications are of great interest in military situations today, and we expect that interest to grow in industrial and eventually consumer environments as well. Mobile agents are one way to realize such applications, especially when used in a wireless environment.
See caripe:awareness for more information.
Mobile agents have received much attention recently as a way to efficiently access distributed resources in a low bandwidth network. Planning allows mobile agents to make the best use of the available resources. This thesis studies several planning problems that arise in mobile agent information retrieval and data-mining applications. The general description of the planning problems is as follows: We are given sites at which a certain task might be successfully performed. Each site has an independent probability of success associated with it. Visiting a site and trying the task there requires time (or some other cost matrix) regardless of whether the task is completed successfully or not. Latencies between sites, that is, the travel time between those two sites also have to be taken into account. If the task is successfully completed at a site then the remaining sites need not be visited. The planning problems involve finding the best sequence of sites to be visited, which minimizes the expected time to complete the task. We name the problems Traveling Agent Problems due to their analogy with the Traveling Salesman Problem. This Traveling Agent Problem is $NP$-complete in the general formulation. However, in this thesis a polynomial-time algorithm has been successfully developed to solve the problem by adding a realistic assumption to it. The assumption enforces the fact that the network consists of subnetworks where latencies between machines in the same subnetwork are constant while latencies between machines located in different subnetworks vary. Different versions of the Traveling Agent Problem are considered: (1) single agent problems, (2) multiple agent problems (multiple agents cooperate to complete the same task) and (3) deadline problems (single or multiple agents need to complete a task without violating a deadline constraint at each location in the network). Polynomial and pseudo-polynomial algorithms for these problems have been developed. In addition to the theory and algorithm development for the various Traveling Agent Problems, a planning system that uses these algorithms was implemented.
See moizumi:thesis for more information.
We are exploring optimal observation of large collections of changing objects. These objects can change at random times, so we cannot know the state of objects for times at which they are unobserved. The goal of an observer is to maintain acceptably accurate state estimates while minimizing observational cost. Our goals are to (1) create models for this type of system, (2) show how these models can be empirically constructed by actually observing real systems, and (3) develop efficient algorithms for the optimal allocation of observation resources within this framework. An example of this type of optimization arises in the observation of World Wide Web (WWW) documents by web search engines and related web software applications. Our initial results include (1) developing statistical models for such systems; (2) collection of empirical data about how web documents change; and (3) development of finite-horizon algorithms for maximizing an index's accuracy. Although the algorithms presented are intended for optimizing the recency of document indices, they are general enough to be applied to any dynamic collection of objects. The work is important and novel in that it takes proper account of the cost of observing, a concept that is crucial to monitoring problems in many communications systems.
See brewington:observe for more information.
What is the traffic hauling capacity of wireless networks? We are examining what is the traffic level that wireless networks can provide. Due to the contention for space in the wireless channel, we have shown that the throughput for each user diminishes as the number of nodes increases. This result as well as its implications are described in gupta:connectivity, gupta:capacity, and gupta:routing, as well as in the paper "Internets in the Sky: The Capacity of Three Dimensional Wireless Networks" (gupta:3dcapacity).
What should be the range of transmission in wireless networks? Clearly the range should be large enough so the network is connected, yet not so large that it is wasteful. We determine the critical range at which the network loses or gains connectivity in gupta:connectivity.
In ad-hoc wireless networks, the nodes may be mobile, nodes as well as links may go down, and traffic patterns may change. A proposed adaptive routing protocol is described in gupta:routing.
An account of problems in wireless networks, including power control, congestion control, medium-access control, routing and capacity analysis, is given in "New Technological Vistas for Systems and Control: The Example of Wireless Networks" (kumar:vistas).
We have developed a new medium-access control algorithm called "Seedex". It is described in kumar:vistas.