Dirección de Comunicación Social del Gobierno del estado de Durango

Stathopoulos et al. proposed an application based collision avoidance for wireless

sensor networks [37]. Three different mechanisms have been proposed in this work.

The first one uniformed TCP-like collision avoidance suggests an additive increase and multiplicative decrease AIMD based on NACK. For each successful packet transmission the application will additively increase the transmission rate. When a NACK packet is received the rate is decreased. In the second informed TCP-like collision avoidance, a similar AIMD scheme is suggested. On the contrary, in this mechanism packet failure information such as collision or link loss are added in the NACK packet. The source will decrease its packet transmission rate only if the packet loss was caused by a collision. Finally a phase-offset collision avoidance mechanism is proposed. This mechanism incorporates a time offset indicating the largest silent period in the NACK packet. Every time the source receives a NACK, a non activity timer is set based on the time offset.

In [38], authors have proposed the Simple, the Adaptive and the Range Ad- aptive Backoff Protocol SRBP, ARBP and RARBP schemes for efficient collision avoidance. All three protocols operate in a similar manner and calculate a back off time for packet retransmission when a collision occurs. In SRBP the back off period is selected uniformly randomly from a continuous space of numbers. This space of numbers is pre-configured. The ARBP protocol is based on the assump- tion that parameters such as network density and packet transmission rate are known in advance and produces a space of numbers based on these parameters. RARBP protocol adjusts the back off time based on the distance between the sender and the receiver of the message.

A hybrid collision avoidance method is proposed in [39]. In this study each

node operates in two alternative modes, sender initiated (SI) and receiver initiated (RI). SI is the default mode, and uses a four-way RTS-CTS, data, ACK handshake. RI is the newly introduced mode in this hybrid mechanism, which operates with a three-way collision avoidance handshake: i) request for request to send (RRTS), ii) multiple access with collision avoidance by introducing (MCA-IB) and iii) col- lision free receiver initiated multiple access (RIMA). A node switches to RI mode only when it does not perform well in SI mode mode. In order to perform receiver initiated handshake, both sender and receiver need to enter RI mode. By ad- aptively sharing the burden of collision avoidance hand shake between the nodes, better fairness and higher throughput is achieved.

In [40], the authors study carrier sense performance and demonstrate that it

can significantly improve the performance in heavy traffic conditions. However, carrier sense has some limitations, originating from the fact that the sender relies on local information to predict the packet reception probability. This can result in lack of information related to parent nodes, which in turn can cause collisions and thus low channel utilisation.

In idle sense [41], each node observes the number of idle slots between two

transmission attempts and compares their theoretical estimate. It then adjusts the contention window via an Additive Increase and Multiplicative Decrease (AIMD) algorithm. In reality, idle sense is a modification of CSMA/CA; after contention, nodes dynamically converge (in a fully distributed manner) to similar values of their contention window instead of relying on exponential back off. In order to achieve this, a relation between the current state of the network and controlling the contention window is established. Idle sense adjusts the contention window when a collision occurs rather than detecting the collision to control the congestion. Transmission opportunities are allocated to the nodes based on the number of idle slots. This method results in higher throughput and short term fairness.

time, Power back off (CSMA/PB), CSMAPB enhances CSMA by adding a trans- mission power control component called power back off (back off in space compon- ent). The authors argue that backing off in space is more efficient than backing off in time, demonstrating that by reducing transmission range by 50% results in a four fold decrease in contention. Low transmission range leads to a low contention path and thus higher network throughput. Using this method may lead to extra routing message exchanges in order to adjust the routes to the sink, leading to slightly increased network overhead.

Enhanced CSMA [42], attempts to improve the performance of CSMA by

lowering the cost of channel state, by adding a learning approach in order to predict the probability of successful reception. In E-CSMA nodes keep state in- formation about all their neighbours. This is acquired by recording the successful reception probability for each neighbour. Before transmitting a packet, a node uses current channel state and the reserved information of the intended receiver as references.

A contention access method for collision avoidance is suggested in [43]. This

work assumes the presence of a DC MAC protocol. In case of collision, all con- tending sensors reduce their contending probability to half. Based on this, only half of the contending sensors will wake up during the next transmission. This algorithm ensures 50% lower probability for collision each time it is trigered.

In [44], authors have proposed a grant-to-sent approach in order to avoid colli-

sions. Grant-to-send mechanism is implemented as an addition to the traditional CSMA/CA MAC. When a node sends a packet, it also informs other neighbour nodes to remain inactive so they will not collide with the recipient’s future trans- mission. Nodes are also sharing some estimated information about the recipients future actions and thus a higher degree of collision avoidance is achieved.

IBPS: a fault tolerant wireless sensor MAC protocol for efficient collision avoid-

ance is proposed at [45]. When a collision occur, IBPS nodes calculate the back

off packet retransmission time and propagate it to all neighbour nodes. Nodes are then calculating future back off periods based on the information received by their neighbours. This mechanism eliminates the probability of collision but adds significant packet overhead.