Wireless Network Throughput Optimization Based On Probabilistic Error Correction And Packet Loss Feedback

With the widespread usage of mobile devices like cell phones and pads,and in-creasing popularity of applications like video streaming,people have higher demand for data rates in wireless network.However,current wireless network,especially mo-bile network can hardly satisfy such demands,and this poses great challenges in flow rate optimization in wireless network.In wireless network,apart from the link congestion caused by limited bandwidth,the noise effect and signal decay can also cause transmission error and packet loss.Congestion control in such conditions will cause unnecessary decrease on the network throughput.Therefore if we can avoid transmission error and packet loss caused by noise effect,that will improve the throughput.Besides,in traditional TCP congestion control,the connection will detect packet loss and invoke the congestion control to guarantee the transmission quality.However,TCP level congestion control can only attain transmission level fairness,and does not guarantee user-level fairness and there-fore cannot guarantee the overall network performance.In this work,we propose two corresponding innovative techniques to optimize throughput in wireless network.Firstly we investigate the possibility of using error estimating code(EEC)for error correction.In wireless networks,to address the problem of packet corruption during transmission,we normally retransmit the corrupted packet,which leads to high redundancy.However,in several applications like video streaming,partially correct packets can also be used.So we can use EEC to estimate the number of error bits in a packet after transmission,and retransmit packets with high bit error rate.In this paper,we investigate whether we can also use EEC for error correction.If we can use EEC to correct the corrupted packets,at least to some extent,that will further reduce the retransmission and redundancy.We first prove the error correction capability of EEC.Then we propose an efficient error correction algorithm based on the burst-error property.Experimental results on a real WiFi trace shows that over 40%corrupted packets can be corrected by our algorithm.In our second work,we consider the problem of traffic allocation among users based on packet loss feedback.In wireless network traffic allocation,user-level fairness is very important to the network throughput.Consider a network of multiple users,and each user can use multiple interfaces or routes.In this network,each user hopes to get more flow rate,and we want to allocate the flow rates of all users to maximize the utility in the whole network.We model this work as a priority-based utility maximization problem,and propose a distributed algorithm for this problem based on packet loss feedback.We show the effectiveness of our algorithm with theoretical analysis and experimental results.Through our error correction algorithm,we can reduce the transmission error and packet loss caused by noise effect and avoid unnecessary congestion control.Then we use our traffic allocation algorithm based on packet loss feedback to improve the overall performance of the wireless network.