Simulation Setup. As deployment setup, the Vodafone LTE small cell testbed network deployment shown in Fig. 2.2 was used. The test network covers an area of approximately one square kilometre and includes two macro sites and 21 small cells represented by the black shapes and red dots, respectively. We use this existing testbed to simulate a relatively dense HetNet scenario. The propagation model is based on a high resolution 3D ray tracing path loss prediction model. This model takes into account clutter, terrain and building data and it guarantees a realistic and accurate propagation model. The user distribution is based on real traffic data extracted from the live network. We assume an inband operation of D2D where D2D UEs use the UL frequency band assigned for cellular (licensed) transmission. However, D2D and cellular UEs are scheduled on different resources which is termed as ’overlay’ operation. The results are based on Monte Carlo simulations and are averaged over 100 simulation runs. The operating frequency applied for the simulations is 2.6 GHz. The maxi- mum transmit powers of macro-cells, small-cells and UEs are 46, 30 and 23 dBm respectively. The fractional path loss compensation power control algorithm in (3.18) is valued with P0 = −90 dBm and α = 0.8 (considered as an optimal value in [54]. An average number of links of 336 is herein considered. Further, without loss of generality, Ith is set to -130 dB as it is proven that this value cre- ates multiple instances of association ambiguity and is worth investigating. The
Simulation Setup. The simulation setup consists of three separate sub-systems: a) the ASV platform, b) the acoustic localization and c) the diver. The simulation is implemented in the Robot Operating System (ROS) where the diver and ASV have each their own ROS master. The ROS masters communicate through an acoustic modem simulation that corrupts the diver position information with noise and randomly creates dropouts in the communication. Using this setup the concept can be tested and proven using pure simulation experiments. This was already done in (Xxxxxxxx, et al. 2013). However, switching directly from pure simulation to human in-the-loop operation is not advised. Therefore the simulation experiments are refined using the hardware in-the-loop simulations. The experiments are made increasingly complex by introducing new hardware agents instead of simulated ones. In the case of diver leader ASV follower two such iterations are needed:
1. ASV hardware in-the-loop – the diver and acoustic interogation are simulated but the real vehicle is used
2. ROV hardware in-the-loop – the diver is simulated using a remotely operated underwater vehicle. The acoustics is performed in a real-environment but without any diver interference or diver related noise The third iteration would introduce a real diver at which stage the experiments no longer contain simulated parts. This experiment can then be used as validation. Experiments are designed to be repeatable therefore in all scenarios the diver is simulated as moving between two points. This provides an idealized, but easily analyzed, version of the diver movements underwater. On the borders of the transect the diver abruptly switches the course of his movements thus showing the amount of overshoot that can be expected due to slow update rates of the diver estimation filter.
Simulation Setup. The simulation consists of three main agents: a) diver, b) buddy and c) ASV. The ASV is included as the communication infrastructure will have to take into account that two USBL devices and one modem are operating in a predetermined three agent interrogation cycle. Hardware will be integrated into the simulation similar as in section 5.3. Four iterations can be defined:
1. Diver tablet hardware in-the-loop – all systems are simulated, except the diver tablet that is fully operational as if used underwater to validate the diver application
