The SPINDLE project aims at being able to send data "toward" the destination even when there is no complete identifiable path all the way to the destination, something which is not achievable with current TCP/IP networks. A key aspect of the DTN is its ability to opportunistically communicate using episodically or intermittently available links. To accomplish this, it will organize information into bundles rather than packets and will route the bundles through intelligent "custodians" that augment traditional routers. These custodians will be responsible for advancing the bundles to the next node en route to their destination. In some cases, that may require initiating novel methods of advancing the information, such as using unmanned aerial vehicles (UAVs) to advance message traffic when there is an obstacle in the path-whether it be geographic or structural-or in the presence of an enemy threat. The result will be a network that functions in the changing and unpredictable environments where reliable communications are most challenging and most critical.
BBN is participating in the Communications and Networking (C&N) consortium of the ARL/CTA program which has 5 different consortia: Advanced Sensors, Power & Energy, Advanced Decision Architectures, Communications & Networks, and Robotics. My involvement in this program is related to specific research problems:
MIMO technology exploits multipath scattering to achieve high spectral efficiency. The basic idea is that instead of using the RF channel as a single channel with multiple reverberations, we use each individual multipath as a separate channel. To do this one has to code orthogonally across all the channels, which results in ~ 10X increase in capacity. However, this also introduces sigificant processing latencies both at the encoder and decoder, and hence poses many challenges to the MAC protocol. We were subcontacted to build the MAC and assist in the integration of the overall network as a subcontractor to Lucent Technologies who developed the MIMO PHY hardware.
For teams of autonomous robots to fully utilize their distributed problem solving and task distribution capabilities, they require a self-organizing, self-healing ad hoc networking protocols. However a particular advantage of robots over other devices that use ad hoc networking protocols is that the algorithms for networking reside in the same architecture as the algorithms for performing motion and mission taskings. Robots can therefore easily and quickly exchange information between these subsystems in order to enhance and improve the ability of the robotic team to perform a task. Although the networking communication layer has traditionally been perceived as a black-box service, we've shown that providing timely access to the information kept by the ad hoc routing protocols imparts a new "sensory" perception to the robot. Likewise, the ad hoc network benefits from access to information about the environment or future robot motion, since it can then adapt to and optimize for current and future operating conditions. This this program we designed and implemented a networking architecture called ERNI (Exploiting Robotic mission and Network Interactions) which provides a simple, open-source based, shared blackboard environment for presenting volatile networking statistics and information to other modules in a system. We have demonstrated this work in simulation and a live robotic testbed in in order to illustrate increases in efficiency on certain tasks, as well as to allow the autonomous tasking control and behaviours to make better decisions given network-observed information.