In a joint effort with the Communications Research Centre Canada, we explore mobility issues in AdHoc networks. In particular, we are implementing OLSR (Optimized Link State Routing) for IPv6 on Linux laptops with IEEE 802.11b wireless LAN cards. We are also integrating this pro-active routing protocol with Mobile IPv6. The protocol implementation will also be ported to the XORP extensible open router software platform, currently being developed primarily at the International Computer Science Institute in Berkeley. Ultimately we plan to use this routing protocol as basis for future work on QoS in AdHoc Networks. To that effect, some simulation work in Opnet and NS2 explores extensions to OLSR, such as:

QoS OLSR: enhance the routing protocol from a best-effort version to a QoS version. Rather than finding the shortest-hop path, QoS OLSR will find, with high likelyhood, the path with maximum bottleneck bandwidth.
Based on QoS OLSR, can we optimize SWAN, a stateless network model that uses distributed control algorithms to support "soft" real-time services and service differentiation in wireless ad hoc networks.

Other work in routing in AdHoc Networks explores a range of different aspects. As mobile devices and sensors will most likely be powered by battery, energy consumption is an important protocol characteristics. We are currently developing PIES, or Protocol-Independent Energy Savings, a general approach to reducing the energy consumption of routing protocols (as opposed to designing a specific energy-efficient routing protocol).

Another important characteristic of multihop wireless networks is the limited network capacity. These capacity limits are well-known and theoretical papers point out that per-node bandwidth tends towards 0 as the network scales in physical size and number of nodes. Even worse, current routing protocols tend to route data "through the middle" of the network, as they are looking for the shortest path. This will cause congestion in one part of the network, while transmission capabilities in other parts of the network (at "the edge") may be underutilized. We are currently investigating how to use the total available network capacity to its fullest by distributing the load in the network more evenly. There are some important trade-offs: as we route data over non-shortest paths, the number of packet retransmissions and the overhead to maintain these longer routes will increase, reducing the per-node throughput.