Research On Collision Recognition Mechanism For Multi-hop Wireless Sensor Network

Hidden terminal effects in multi-hop wireless sensor networks based on IEEE802.15.4 may cause serious collisions in packet transmission.Compared to non-collision errors caused by path loss,multipath fading,shadow fading,and IEEE 802.11 interference,collision errors caused by hidden terminals are persistent and destructive.Accurate recognition of collisions caused by hidden terminals is thus critical to guarantee the correctness of data transmission.This thesis first studies the mechanism of collisions caused by hidden terminals and the impact on network transmission performance.The complete collision and retransmission process are analyzed by a numerical analysis method.By constructing the communication model,the collision and retransmission probability of the packet in non-beacon-enabled CSMA/CA are obtained.On this basis,the factors affecting network performance are studied.Theoretical analysis and simulation results show that the packet length,packet generation rate,and hidden node ratio all have impacts on network performance.Among them,the packet generation rate is the most important factor.The error characteristics of transmission error packets are then studied through high packet generation rate experiments.The transmission experiments in typical scenarios are designed to obtain the error packets in non-collision error scenarios and collision scenarios.Next the bit-error rate,bit-error distribution rate and symbol-error distribution rate are designed as performance indicators to calculate the error characteristics.The statistical results show that the latter part of the collision packet tends to have a high probability producing errors.This feature clearly distinguishes collision errors from non-collision ones in different scenarios.To utilize the error characteristics of the collision error packet,this thesis further proposes a collision recognition mechanism based on the idea of redundant domain insertion,called CRM-ML.This mechanism inserts the redundant domain into the packets,with pre-defined content known by the receiver.The location of redundant domain is chosen by a mutual-information-inspired technique.The redundant domain in the packet can be classified by a machine learning classifier to determine whether the transmission failure is caused by collision or not.The experimental results show that the CRM-ML mechanism achieves good recognition accuracy over 90% with94% coding efficiency.Built on the proposed CRM-ML mechanism,this thesis finally designs a MAC-oriented hybrid switching mechanism to achieve adaptive switching of the MAC mechanism in the event of consistent collisions.The designed switching mechanismis applied to the CSMA/CA and CDMA hybrid protocol,which verifies the effectiveness of the proposition.The proposed collision recognition mechanism can directly analyze each received error packet at the receiver without storing or collecting extra packets when the wireless sensor network is running.The collision recognition based switching mechanism also provides an effective switching point for the hybrid MAC protocol,thereby,maximizing the network communication performance.