Network Utility Maximization for BATS Code Enabled Multihop Wireless Networks

Network utility maximization (NUM) is studied for multihop wireless networks employing an efficient random linear network coding scheme called BATS codes. Compared with the classical random linear network coding scheme, BATS codes have lower computational and storage costs at the intermediate network nodes, and can achieve close-to-optimal end-to-end throughput and latency for multihop networks with packet loss. We formulate a NUM problem that optimizes the total utility of multiple communication flows under certain link scheduling constraints. Our problem employs a practical throughput measure induced by BATS codes and hence can provide realistic guidelines about network protocol designs for multihop wireless networks. Our problem in general has a non-convex objective function with integer variables, so that the algorithms of solving existing network utility maximization problems cannot be directly applied to our problem. We discuss a modified dual-based algorithm for solving our problem and evaluate its performance numerically.