This page lists some future tasks and research questions that came up while we were planning the demo.
Should we and how do we optimize the network for "military" traffic patterns?
Cluster SALUTE and BGW records. Will the existing clustering algorithms work well with really short text fragments (i.e., the text from the description fields in the SALUTE and BGW records)? How do we extend the clustering algortithms to use the non-text fields such as height and height?
Explore different strategies for getting a message through the wireless network as fast as possible (e.g., send the message along multiple, randomized paths)
The platoon leader could have automated tools that scan the data coming from the soldiers and detect problems (e.g., are all the soldiers in the right place). The platoon leader would have a database of soldier information against which these sentinel agents (or persistent queries) would work. These sentinel agents could also use various kinds of network information, so they tie in well with future network-sensing work.
The platoon has a platoon leader who is responsible for filtering the SALUTE reports that flow from the platoon back to Battalion HQ, and the confirmation requests and order updates that flow from Battalion HQ to the platoon. What do you do if the platoon leader is incapacitated, or the platoon leader's computer breaks? In the first case, computer control must be transferred to the next person in the platoon's chain-of-command. In the second case, the platoon leader might take someone else's computer, or might delegate computer control to another platoon member. In both cases, a computer must re-configure itself as the platoon leader's computer.
Use MACE to generate a mission plan. An agent carries the mission plan onto the soldiers' machines, guides them through the machine, and monitors mission status.
A group of soldiers might move out of network range of the rest of the soldiers, creating a disconnected cluster of machines. Do you and how do you elect a temporary (computer) leader for such a disconnected cluster?
If a group of machines moves away from the main network, Lisa's current software will not detect that these machines are about to become disconnected from the network, since the machines will still have network connections to each other (Lisa's current software detects if a machine is losing all its network connections). So (1) Lisa's software must detect individual link failures, (2) information about individual link failures must be propagated through all machines in the local cluster so that the machines realize that their local cluster is disconnected or about to become disconnected (Aprl 2 gives a reasonable approximation of the disconnected part, since only reachable machines will remain in the routing table, but Aprl 2 can not predict impending link failure), and (3) the system must tell soldiers about impending link failure, tell them what to do about it, etc. For example, we could flash network links that are about to go down in red, tell the soldier where they could move physically to main network connectivity, etc.
We need a more complete networking solution. For example, in some environments, the soldiers might be able to set up a huge antenna (i.e., the equivalent of a cellular-phone tower). A routing protocol that is terrible or impossible when there is no antenna might then become the best choice.
Based on the mission plan, the system could automatically launch queries. For example, if the platoon is getting close to the point where they will move to the building and secure it, the system could automatically launch queries about the building and its surrounding.
Memory-based retrieval, i.e., search for similar missions and see how those missions turned out.