Sustentabilidad; donde el punto central es la gente

For D2D pairs to determine a suitable transmission mode to operate in, they base their decision on two factors, i.e., the measured performance gain they will receive and information about the network. Typically, the BS will assist D2D users in determining a particular transmission mode to operate in, which is referred to as network-assisted D2D communications [16]. Thus, the BS will help D2D users by providing additional network information, as well as coordinating the D2D pairs and cellular users within the network to reduce severe intra-cell interference. The type of network information that the BS will provide to the D2D users will include, the physical distance between the D2D pair transmitter and receiver, and the complete CSI, as well as the interference that will be received. Moreover, mode selection can also be based on the D2D pair’s performance, i.e., the D2D pair will prefer to operate in a particular transmission mode if its throughput is increased or its energy consumption is reduced.

Within literature, there is a wide range of existing D2D mode selection tech- niques, such as in [16, 33, 84–88] and [89]. Distance dependant mode selection is one of the simplest mode selection approaches, where the distance between the D2D pair transmitter and receiver must be within a threshold distance to transmit directly to each other, otherwise the D2D pair will operate like traditional cellular users. The work in [84] proposes a mode selection approach based on a threshold distance in order to minimise transmission power. The proposed mode selection approach in [84], found that the distance threshold was inversely proportional to the BSs density, as well as being strictly increasing with the path loss exponent. Meanwhile, the work in [86] considers three factors for mode selection which are, the distance between the D2D pair transmitter and receiver, link quality, and a bias factor. Thus, the proposed approach in [86] enables D2D pairs to only trans- mit directly between one another if their channel link quality is at least the same quality as the cellular user uplink. On the other hand, in [88] a mode selection

2.1 Resource Management and End-User Satisfaction in D2D Communications 31

technique based on D2D and cellular link quality, and interference was proposed. In particular, the interference considered in the mode selection approach in [88] also factors in the intra-cell and inter-cell interference. In contrast, the work in [89] considers a mode selection approach based on optimising the sum rate of D2D communications, and takes into account the cellular, dedicated, and reuse modes.

As outlined in Section 1.3.2, game theory is a powerful mathematical tool that enables decision making and models the interactions between users. Game theoretic techniques have been considered for dynamic mode selection, such as in [77,90–95], and [96], to model the interactions between D2D pairs wishing to change their transmission mode. In the aforementioned works, the mode selection problem has been formation using a two-armed bandit game, a coalitional game, or an evolu- tionary game. The work in [90] proposed a two-armed Levy bandit game for D2D mode selection, where mode selection is based on maximising an expected reward, which is a function of the strategy the D2D pair chooses and its utility. On the other hand, the work in [93] proposes a coalitional game for energy-efficient mode selection, where D2D pairs in the same transmission mode can cooperate and share information. Thus, in [93], the energy-efficient mode selection approach focuses on minimising transmission power while guaranteeing rate requirements. Meanwhile, the work in [96] proposed an evolutionary game for D2D mode selection, where replicator dynamics are used to model the preferences of D2D pair’s transmission mode based on the utility function. Therefore, game theoretic approaches enable D2D pairs to dynamically change their transmission mode, only if their payoff is increased.

In fact, existing works, such as in [91–95, 97–99] consider a distributed mode selection scheme, however, D2D pairs have only been considered to operate in either reuse mode or cellular mode. In addition, these works only consider one D2D pair to reuse only one or more cellular user’s resource block. However, in practice, these approaches are not ideal, as spectral efficiency and network performance cannot be maximised due to limiting the number of D2D pairs reusing cellular user resources. A limited number of works, such as [36, 93], and [100] have considered D2D pairs operating in either reuse mode, cellular mode, or dedicated mode. In particular, when considering dedicated mode, coordination is required between D2D pairs and the BS, as the BS needs to assess if there are any free resource blocks available to be allocated to D2D pairs [36]. This coordination can cause signalling overhead to

32 Literature Review

increase, however, D2D pairs operating in dedicated mode will have ultra-reliable communications, which, in turn, will improve overall network performance and throughput.

In most of the works described thus far implement network-assisted D2D com- munications, where the BS assists D2D users in determining which transmission mode to operate in. Thus, in network-assisted D2D communications, it is assumed that the BS must have global knowledge of the network, such as, complete CSI and the location of all users [16,31–33]. In order for the BS to acquire global knowledge of the network, this will cause signalling overhead and network complexity to sig- nificantly increase, especially with the increasing number of users in the network. Furthermore, in practical scenarios, network-assisted D2D communications is not ideal, as complete CSI cannot always be achieved for all users all the time.

Hence, as the network begins to evolve towards multimedia IoT, users will start to become more self-sufficient and intelligent. Thus, the BS won’t need to provide information to users all the time, i.e., the BS will not be required to have full knowledge of the network. The work in [36] considers a D2D cognitive communi- cation system with assistance-free D2D mode selection, i.e., assuming partial CSI, that analyses the network performance and and the impact of direct transmission between D2D users on the cellular network. Whereas [100] jointly considers mode selection, user scheduling, and rate adaptation, with partial CSI, to ensure user fairness and reduced interference in the network. Even though some of the pro- posed works, such as, [36,101], and [100], consider limited/partial CSI, they do not jointly consider mode selection, resource allocation, and interference management for D2D communications in a distributed and autonomous manner.