The goal of this task, which is part of a larger ONR-funded project at BBN Technologies is to address the problem of routing to terminal nodes(endpoints) with very high "effective" mobility i.e: when the period between changes in an endpoint's network attachment point is comparable to the time for the location tracking mechanism to converge.
This could happen due to higher endpoint speed, smaller cells, or increased control messages latency. When this happens, conventional reactive location tracking mechanisms fail.
That is, by the time a data packet is forwarded to the last "known" location, the endpoint has already gone elsewhere.
 Anticipating future network's attachment points and forwarding data packets directly to those locations may be required to reduce the number of interruptions in sessions involving those endpoints.  Three techniques have been developed and implemented to address this issue:

• Spray Routing:   Spray routing multicasts data traffic within the vicinity of the last known location. It is the most appropriate when no trajectory information is available, but the last location of the endpoint is known. "Spraying" the switches in the vicinity of the last known location of the destination should ensure that packets will be delivered to the moving endpoint. This even if it moves and affiliates with a neighboring switch during the packet transit time.

• Trajectory Routing: Trajectory routing estimates where a session's destination endpoint will reside at a particular time and forwards data traffic directly to that location. This is a predictive mechanism in the sense that it uses past trajectory information in order to estimate future network's attachment points along with estimated arrival times to and departure times from those locations. Unlike a "Spray" mechanism, we assume switches  to be GPS-capable in our current implementation.

• "Trajectory with Spray" Routing: This is a "unified" and adaptive scheme that resorts to a reactive mechanism when the mobility is low. Future locations are predicted for high mobility and data packets are "multicast"(sprayed) within the vicinity of the selected and predicted location.

Our discrete event simulator has been built from scratch using C++.  This after careful consideration of several simulation environments including OPNET, NS-2, parsec and so on. This allows us to independently control the tradeoff between fidelity(detail) and performance at each layer of the stack. In our simulator, mechanisms that have a large impact on the main thrust of the study are modelled in greater detail, and those that are necessary but have a minor impact are abstracted. This allows us to run a set of large scale simulations to get the confidence we desire while not losing touch with the reality.

All the work has been done under  FreeBSD version 3.4. Highlights of our code that can be downloaded along with general instructions include Reactive Location Tracking Scheme,  Spray Routing , Trajectory Routing  and the Unified Scheme.  Our reactive scheme has been built as the baseline of the study of routing to highly mobile endpoints. (see  paper for description).  Spray routing, Trajectory routing and the unified scheme are also described in paper

•     F. Tchakountio, R. Ramanathan ,  "Tracking Highly Mobile Endpoints",   In proc. of ACM Workshop on Wireless Mobile Multimedia (WoWMoM01), July 2001, Rome, Italy.
•     F. Tchakountio, R. Ramanathan ,  "Routing to Highly Mobile Endpoints: A theoretical and Experimental Study of Limits and Solutions" ,  BBN Technical Report no. 1280 , December 2000.