Skip to Main Content
The use of traditional TCP, in its present form, for reliable transport over ad hoc wireless networks (AWNs) leads to significant degradation in the network performance, in terms of reduction in average network throughput and increase in packet losses. This is primarily due to the congestion window updation and congestion control mechanisms employed by TCP. TCP follows a deterministic approach for updating the size of the congestion window, which is less suitable for loss-prone AWNs as it leads to very frequent occurrences of congestion in the network. Another reason is TCP invokes the congestion control mechanism for both congestion and wireless losses, as it cannot distinguish between them. Hence, in order to use TCP in AWNs, an efficient mechanism must be provided with TCP for updating the size of the congestion window based on the network conditions and distinguishing the congestion losses from wireless losses. In order to address these problems, we propose Learning TCP, a novel learning automata based reliable transport protocol for AWNs, which efficiently adjusts the size of the congestion window and thus reduces the packet losses. The key idea behind Learning-TCP is that, it dynamically adapts to the changing network conditions by observing the occurrence of events, such as arrival of acknowledgment (ACK) and duplicate ACK (DUPACK) packets and appropriately updates the congestion window size. In addition to this, we use a deterministic approach for packet loss discrimination, in order to take the appropriate action for each type of loss. Learning-TCP, unlike other existing proposals for reliable transport over AWNs, does not require any explicit feedback, such as congestion, link failure, and available bandwidth notifications, from the network. We provide extensive simulation studies of Learning-TCP under varying network conditions that show increased throughput and reduced packet loss compared to that of the traditional TCP.