Chapter 6: An ACO Algorithm for Data Network Routing

6.1 The Routing Problem

• Networks can be circuit-switched or packet-switched
• Each node typically builds a routing table
• We will consider fluctuating traffic, but not fluctuating network topology
• Routing can be centralized or distributed
• Routing can be static or adaptive
• The focus can be optimal routing or shortest path routing
• Shortest path routing can use a distance vector algorithm (such as Bellman-Ford) or a link-state algorithm
• A distance vector consists of a destination, an estimated distance and a next hop

Communication Network Model

• Hierarchically organized WANs. Subnetworks have a gateway node. Groups of gateway nodes have an area border gateway.
• Each node has a buffer space for incoming and outgoing messages
• The buffer space is divided between low priority (for data packets) and high priority queues (for routing packets)
• Once sent, a packet is considered to be acknowledged
• Links between nodes have a bandwidth and a transmission delay
• The network is simulated

6.2 The AntNet Algorithm

The Algorithm

• AntNet Algorithm: page 232

Data Structures

• Τ_i: τ_ijd where i the current node, j is the next node to move to and d is the destination
• Μ_i (μ_id, σ²_id, W_id) where W is a moving window that stores the recent best trips
• Figure 6.1

Data Structures Update

• Figure 6.3