Distributed Satellite-Terrestrial Cooperative Routing Strategy Based on Minimum Hop-Count Analysis in Mega LEO Satellite Constellation

Our study primarily focuses on the mega Walker Delta constellation, which comprises $N$ orbits with an acute orbit inclination $\alpha$, and each orbit contains $M$ satellites. As illustrated in Fig. 1, $N$ orbits are uniformly distributed in the equatorial plane, and the difference between adjacent orbits in the longitude of ascending node is $\Delta\Omega=2\pi/N$. Moreover, $M$ satellites are also uniformly distributed in each orbit, and the phase difference between neighboring satellites within an orbit is $\Delta\Phi=2\pi/M$. Meanwhile, the phase difference between two satellites connected by an inter-orbit link is $\Delta f=2\pi F/(NM)$, where $F$ is phasing factor given as a constant in the design of constellation. In this paper, each satellite can establish four permanent ISLs with its neighboring satellites, consisting of two intra-orbit links and two inter-orbit links, and all satellites are further categorized as ascending satellites and descending satellites[33] based on their moving directions. Ascending satellites are moving in the latitude-increasing direction, while descending satellites are moving in the latitude-decreasing direction.

The introduced satellite-terrestrial cooperative routing framework is shown in Fig. 2, which consists of ground nodes and a mega Walker Delta constellation. Here, ground nodes include source nodes, destination nodes, and ground relay nodes. The source and destination nodes are accessed into the constellation as users[43]. Every ground relay node can establish several SGLs with satellites within elevation range and make routing decisions according to routing strategy[44, 45]. $GR_{k}$ is used to denote the ground relay whose ID is $k$. Moreover, all satellites in constellation possess on-board processing capability, enabling them to forward packets to neighboring nodes via ISLs or SGLs. In more detail, satellites that provide access service to source and destination nodes are defined as access satellite, which are represented as $S_{c}$ and $S_{d}$, respectively. Satellites with SGLs are referred to satellite gateway, and packets can be forwarded between satellite gateways and ground relays via SGLs. For convenience, we denote the $i$-th satellite gateway belonging to ground relay $k$ as $SG_{k,i}$, and refer to all satellite gateways in system as key nodes. In addition, the selection of access satellites by source and destination nodes is outside the scope of our study, so the end-to-end routing in the following means the routing process from $S_{c}$ to $S_{d}$.

Under the satellite-terrestrial cooperative routing framework, packets can be forwarded in three potential modes, which include inter-satellite forwarding, satellite-terrestrial forwarding, and terrestrial-satellite forwarding. The last two modes represent the forwarding of packets between satellite gateway and ground relay via downlink SGL or uplink SGL, while the inter-satellite forwarding means packets are routed between satellites via ISLs. In this paper, each forwarding via ISL or SGL is defined as one hop, and the inter-satellite routing is performed based on the inter-satellite minimum hop-count estimation results, which specify the routing direction, as well as the minimum number of routing hops required in the horizontal and vertical direction. In Walker Delta constellation, all potential routing directions can be categorized into four types, which involve top-right, bottom-right, top-left, and bottom-left[33]. Taking Fig. 2 as an example, the minimum hop-count $H_{\text{min}}$ from $S_{c}$ to $S_{k,1}$ is 3, with a routing direction of top-right, where the minimum routing hops required in the horizontal and vertical direction are 1 and 2, respectively. Furthermore, in the inter-satellite routing mode with minimum hop-count constrains, each satellite can forward packets to the next-hop satellite in horizontal or vertical direction via ISLs, but when the remaining hop-count in one forwarding direction is 0, packets can only be forwarded to next-hop satellite corresponds to the other direction.

Based on the framework described here, we deeply study the low-complexity minimum end-to-end hop-count estimation method as well as the satellite-terrestrial cooperative routing strategy, and the entire paper is organized as Fig. 3. We first introduce the satellite real-time position based graph (RTPG) to describe 3D constellations in Section III. By using RTPG, the minimum end-to-end hop-count estimation method (KNBG-MHCE) is designed in Section IV. Utilizing the minimum end-to-end hop-count constraints given by KNBG-MHCE and jointly considering the load status of satellites, we propose the distributed cooperative routing strategy in Section V.

In walker Delta constellation, since $N$ orbits are uniformly distributed along the equator at an interval of $\Delta\Omega$, we divide the earth’s surface into $N$ regions along the horizontal direction from $\lambda=0^{\circ}$, and all regions have a width of $\Delta\Omega$. The longitudes corresponding to the left and right boundaries of region $P$ are given by (1). Moreover, satellites are uniformly distributed within their orbits with phase difference $\Delta\Phi$. Therefore, each orbit can be divided into $M$ regions starting from the phase $\frac{\pi}{2}$. However, due to the non-linear relationship between the phase difference and latitude difference caused by orbit inclination, these regions will translate into $M$ unevenly distributed regions in vertical direction of the earth’s surface. The latitudes corresponding to the lower and upper boundaries of region $R$ in vertical direction are given by (2). In the following two equations, $P$ takes value from $0,1,\dots,N-1$, and $R$ takes values from $0,1,\dots,M-1$.

By dividing the regions both horizontally and vertically, the entire earth’s surface is divided into $N\times M$ regions. Referring to Fig. 4, when $N=5$ and $M=10$, the earth’s surface is divided into 5 regions in horizontal direction and 10 regions in vertical direction, and each region can be represented by its coordinates $(P,R)$. In addition, there is only one satellite within each region at any given time.

The coordinates of region where each satellite is located at time $t$ need to be calculated, which is the key to convert the three-dimensional constellation into a two-dimensional graph. Although the position of satellites and orbits in relation to the earth is continuously changing, satellite phase and the longitude of ascending node can be used to indicate the position of a satellite, and determine which region the satellite is located in. For the $m$-th satellite in orbit $n$ at time $t$, the latitude $\varphi_{m,t}$ and longitude $\lambda_{m,t}$ can be obtained according to the ephemeris information. Employing these two parameters, the phase of satellite $m$ and the longitude of ascending node in orbit $n$ can be calculated by (3) and (4), respectively[33].

$u_{m,t}\in[-\pi,\pi]$ represents the phase of satellite $m$ at time $t$. When the phase of satellite is within $[-\pi/2,\pi/2]$, it signifies that this satellite is an ascending satellite. Otherwise, this satellite is a descending satellite. On the other hand, $L_{n,t}\in[0,2\pi]$ represents the longitude of ascending node in orbit $n$. $\mathbb{N}(x)=\text{mod}(x,2\pi)$ in (4) is a normalisation function that ensures the value of $L_{n,t}$ is within the interval $[0,2\pi]$. Based on $u_{m,t}$ and $L_{n,t}$, we can calculate coordinates $(P_{n,t},R_{m,t})$ of the region where satellite $m$ in orbit $n$ is located at time $t$, and $P_{n,t}$ is expressed as

In equation (6), $P_{n,t}$ takes values from $0,1,\cdots,N-1$, and a different value of $P_{n,t}$ indicates that the satellite is located in different horizontal region. Since $L_{n,t}$ will be changed due to the earth rotation, $P_{n,t}$ is a periodic variable with period $T_{e}=2\pi/\omega_{e}$, where $\omega_{e}$ is an angular velocity of the earth rotation. The other coordinate $R_{m,t}$ is given as

Following (7) and (8), $R_{m,t}$ takes value from $0,1,\cdots,M-1$, and a different value of $R_{m,t}$ indicates the satellite is located in different vertical region. Since all satellites move along the orbit with angular velocity $\omega_{s}$, $R_{m,t}$ is also a periodic variable with period $T_{s}=2\pi/\omega_{s}$. Moreover, the phases corresponding to the lower and upper boundaries of region $R$ in vertical direction are expressed as

By calculating coordinates of the region in which each satellite is positioned, we can transform the constellation into a two-dimensional graph model RTPG as shown in Fig. 4.

Before designing KNBG-MHCE method, we need to study the minimum end-to-end hop-count estimation between two satellites under inter-satellite forwarding mode at first. The estimation results will then serve as weights for the edges connecting key nodes within KNBG, and give the routing direction and minimum hop-count constraints for inter-satellite routing as well. Aiming at reducing the computational overhead, a minimum end-to-end hop-count estimation method with complexity $O(1)$ for inter-satellite forwarding mode is designed based on satellite coordinates in RTPG. For brevity, we assume the coordinates of two satellite nodes in RTPG are $(P_{n1,t},R_{m1,t})$ and $(P_{n2,t},R_{m2,t})$, and the minimum end-to-end hop-count between them in inter-satellite forwarding mode is $H_{\text{min}}$. Since the following derivations are applicable at any given time, variable $t$ is omitted.

As illustrated in Section II-B, there are four possible routing directions in inter-satellite forwarding mode. Hence, there will be a minimum end-to-end hop-count in each of these four directions, and $H_{\text{min}}$ is the smallest one of these four values, which can be expressed as follows[33]

In equation (10), $H_{h}^{\rightarrow}$ and $H_{h}^{\leftarrow}$ denote the minimum forwarding hops required horizontally to the right and left, and $H_{v}^{\rightarrow\uparrow}$ and $H_{v}^{\rightarrow\downarrow}$ indicate the minimum hop-count required upward and downward when the routing direction is rightward in horizontal direction. Similarly, $H_{v}^{\leftarrow\uparrow}$ and $H_{v}^{\leftarrow\downarrow}$ signify the minimum hop-count required upward and downward when the routing direction is leftward in the horizontal. With regards to obtain $H_{\text{min}}$, $H_{h}^{\rightarrow}$ and $H_{h}^{\leftarrow}$ are first calculated as follows

Second, (12) and (13) are functions describing the change in coordinate $R$ caused by the phase difference $\Delta f$ when forwarding horizontally to the right and left, respectively.

Substituting (12) and (13) into (14a) results in $H_{v}^{\rightarrow\uparrow}$ and $H_{v}^{\leftarrow\uparrow}$. In a similar way, substituting these two equations into (14b) yields $H_{v}^{\rightarrow\downarrow}$ and $H_{v}^{\leftarrow\downarrow}$.

Finally, based on derivations above, the minimum end-to-end hop-count $H_{\text{min}}$ and the corresponding routing direction for inter-satellite forwarding mode can be obtained according to (10), and the calculation process is summarized in Algorithm 1, which exhibits $O(1)$ complexity.

When one user accesses any satellite, this satellite is going to estimate the minimum end-to-end hop-count for these packets, and the estimation results will provide a minimum hop-count constraint for subsequent satellite-terrestrial cooperative routing. To realize this estimation, it is first necessary to periodically generate the key node based graph KNBG in every satellite, where key nodes refer to all satellite gateways. The generation process of KNBG can be divided into two steps, which contain key node extraction and weighted edge generation. Due to satellite movement, satellite gateways in the system are not fixed, and thus KNBG face frequent regenerations. Therefore, we fully design above two steps to ensure the low complexity regeneration of KNBG. Next, taking KNBG as input, Dijkstra is executed to estimate the minimum end-to-end hop-count. We refer to the method formed by above three steps as KNBG-MHCE, and a more comprehensive symbol description is provided in Section II-B.