Cellular Simulation for Distributed Sensing over Complex Terrains

Abstract

Long-range radio transmissions open new sensor application fields, in particular for environment monitoring. For example, the LoRa radio protocol enables connecting remote sensors at a distance as long as ten kilometers in a line-of-sight. However, the large area covered also brings several difficulties, such as the placement of sensing devices in regards to topology in geography, or the variability of communication latency. Sensing the environment also carries constraints related to the interest of sensing points in relation to a physical phenomenon. Thus, criteria for designs are evolving a lot from the existing methods, especially in complex terrains. This article describes simulation techniques based on geography analysis to compute long-range radio coverages and radio characteristics in these situations. As radio propagation is just a particular case of physical phenomena, it is shown how a unified approach also allows for characterizing the behavior of potential physical risks. The case of heavy rainfall and flooding is investigated. Geography analysis is achieved using segmentation tools to produce cellular systems which are in turn translated into code for high-performance computations. The paper provides results from practical complex terrain experiments using LoRa, which confirm the accuracy of the simulation, and scheduling characteristics for sample networks. Performance tables are produced for these simulations on current Graphics Processing Units (GPUs).

Keywords: LoRa; cellular automata; complex terrain; parallel processing; radio signal propagation.

Figure A1

Heavy flooding in the center of Morlaix, France due to a storm on 3 June 2018 (Photo: Le Télégramme).

Figure A2

Simulation results for Morlaix, obtained by executing massively parallel processes on GPUs with thousands of CUDA cores. The experimental zone is comprised of 58,275 cells corresponding to actual area 3 × 3 km. A place of concern is indicated by a red circle. This is rue de Brest near the river Queffleuth (longitude: −3.8329839706421, latitude: 48.573767695437, elevation: 10.1 m) where the water level went up to 60 cm according to Le Télégramme newspaper. The main flooding directions can be seen in light blue over the maps.

Figure A3

Chart shows rainfall levels over 32 h from 12:00 a.m. on 3 June 2018, to 8:00 a.m. on 4 June 2018. It was noticed that serious flooding often comes several hours after a storm or heavy rain. After the rain stopped, it required a couple of hours or days to drain all of the water.

Figure A4

Even though the heavy rain just occurred during a few hours (3:00–5:00 p.m.), simulation results show that there are 50 points where water levels are from 30 to 70 cm.

Figure A5

A base station is equipped with a LoRa board (emitter) and a MacBook. This base station was deployed on the top of the Roc’h Trevezel mountain, France. The receiver is another LoRa board mounted on a car moving around.

Figure A6

Terrain complexity analysis for the Roc’h Trevezel using TPI metrics. This shows remarkable landform splits in the North–South direction. This provides critical references for network deployment aiming at enhancing the system robustness and range and reducing the cost as well.

Figure A7

The base station (emitter) is located on top of the mountain (latitude: 48.4051306, longitude: −3.9077762, elevation: 345 m). A receiver is placed on a car (receiver) traveling around the area with many hills, valleys, and big trees. From simulation results, places (cells) are able to receive the radio signal highlighted in yellow. Blue circles show points where the receiver successfully obtained messages in our actual measurement.

Figure A8

Obtained received signal strength indicator (RSSI) values for experiment in our actual measurement at the Roc’h Trevezel. RSSI values decay as a negative exponential function of distance. The complex terrains in this area seriously impact the quality of radio links.

Figure A9

Chart of radio coverage execution times for different resolutions of cell from 5 to 30 pixels. This statistics was recorded for the performance on a PC equipped Core i7-7700K CPU@4.20GHz × 8, 16 GiB DDRAM, card NVidia GeForce GTX 1070 (1920 CUDA cores).

Figure 1

Federation of two complementary cellular subsystems, a sensor network, and some display for observation. A software bus (RTI) provides services for data exchanges, sequencing and synchronization. This framework is called High Level Architecture, standardized as IEEE Std 1516–2000 [5].

Figure 2

TPI terrain complexity for the river Soummam in Algeria (see also Figure 5). Red lines represent higher points, difficult to overcome and blue lines are for lower points, difficult to reach. The white zone signals a flat ground without remarkable obstacles, the case of Soummam banks. This grid is 262 × 226 points, representing 30 × 25 km.

Figure 3

A histogram of topography ruggedness. Four zone analysis with high variability (red, the Soummam), low variability (blue, the City of Brest, with several deep valleys and the shore), a zone with an high percentage of sea surface (green, Brest bay), and medium variability with low size hills (brown, the Arrée mountains).

Figure 4

A subsystem of low terrain complexity was extracted from Figure 2. This is a flat zone around the river Soummam, with complexity below the threshold line on the left in Figure 3.

Figure 5

Display of a radio coverage for an emitter (red point), located above the Soummam (36.622141028, 4.799995422), elev:165.0 m. Note that geographic positions are proposed in (lat, long) form compatible with familiar map navigators. Coverage is shown in dark yellow, spreading over 30 km on the zone width. This location was chosen at random, giving a percentage of 32% grid points receiving the long-range signal.

Figure 6

Chart of rainfalls and water levels at two positions in the Soummam river zone during a tropical storm from 13 to 16 November 2017. The blue line shows rainfalls during four days recorded every 3 h [30]. The two lines in red, green color present water level at P1, P2, respectively (also see Table 1). The distance between P1 and P2 is 300 m. Due to the significant slope of the ground surface in this complex terrain, a large amount of water was accumulated at lower points, causing a flash flood, and possibly a catastrophic landslide.

Figure 7

Heavy rainfalls context: a physical simulation produced positions with the risk of flash flooding (dark blue color). The communication coverage of a base station with a star network is predicted for sensor nodes monitoring level of water in these positions. An algorithm selects reachable sensor positions from a network sink, sorts, then extracts 15 positions according to the flooding result simulation. More details about accuracy are given in Section 3.

Figure 8

Quickmap tool [33] showing tile coverage. For this case, zoom factor is 9, and the last tile at the bottom right has x = 251, y = 177 indexes.

Figure 9

Tile map server architecture: tiles are requested from the web page server that either return a cached image or ask the rendering engine to compose it from a data base of geographical objects.

Figure 10

Presentation of a cell system organization over the tiles of Figure 8. Each cell has an identity produced from its location inside the window, and a geographical location. The text window bottom right also displays a plus parameter for the elevation. The cell size is $25\times 25$ pixels, representing 7644 m.

Figure 11

Cell synthesis flow: (1) a zone was located from a Quickmap navigation, and (2) was segmented into cells, then a subsystem was extracted filtering cells with elevations less than 45 m. (3) A cell system was generated following Moore topology. Annotations show controls for geographical positions with a cell size of $5\times 5$ pixels representing a $191\times 191$ m${}^2$ surface (A), classification was operated for elevation (B). The neighborhood was Moore, radius 1 (C).

Figure 12

Cell node representation: (a) is the internal architecture with an automaton (rule) operating on incoming values and local stimuli (stim), based on a local set of variables, filling output communication buffers. Communications (com) are operated to and from the input and output buffers; (b) is the external point of view that only shows bidirectional links of a cell node.

Figure 13

Physical exchange during a rain episode. An incomplete neighborhood from a system shows a center cell with 2 neighbors west and east, and 1 in the south. The physical behavior is water flowing downward, represented here by synchronous messages sending water quantity west to center, and center to the east. Refer to [49] for more realistic behaviors.

Figure 14

Radio signals propagate in concentric squares step by step. Reachable cells are represented in colored stripes.

Figure 15

A profile obtained along a route. Let’s assume that an emitter located on the furthest left of the chart, with a distance of 0 m, an elevation of 350 m. According to LoS condition, three points at distances 2997 m, 3269 m and 3541 m seem unable to receive the signal from the emitter.

Figure 16

Directed BSF is a distributed parallel algorithm able to manipulate data on grid cells. Its aim is to mimic point-to-point radio links adopting the LoS condition. Segmented lines are laid on a map to represent how cells forward incoming signal gradually from a root cell (x = 10, y = 1) to definite directions.