Agent-based Dynamic Routing in the
Packet-Switched LEO Satellite Networks
Zhaofeng Wu
∗
, Guyu Hu
∗
, Fenglin Jin
∗
, Bo Jiang
∗
, Yinjin Fu
∗
∗
College of Command Information Systems, PLA University of Science and Technology
Nanjing, P.R.China
Email: qqk20080915@163.com, huguyu@189.cn, fljin@sina.com, wbjcy54321@126.com, nanyinghnjm@163.com
Abstract—This paper proposes and evaluates the agent-based
dynamic routing (ADR) in the packet-switched low earth orbit
(LEO) satellite networks. There are two kinds of agents, roaming
agents and fixed agents. Roaming agents walk randomly around
the network to gather the latest link status information of
the traversed satellites and at the same time transfer the link
status information of the past satellites to the fixed agents
of the satellites. Meanwhile, the fixed agents located at the
satellites estimate the link cost of the network based on the link
status information obtained from the roaming agents and take
advantage of the predictable property of the satellite networks
to update the routing tables using the HALO algorithm already
proposed for the terrestrial networks. Through simulations on an
Iridium-like satellite network, the proposed routing mechanism
is seen to be able to achieve a balanced traffic distribution across
the satellite network, and can especially decrease the packet drop
rate, provide better throughput and end-to-end delay bound, thus
improve the utilization efficiency of the satellite networks.
Keywords—LEO satellite networks, dynamic routing, agent, load
balancing.
I. INTRODUCTION
The packet-switched low earth orbit (LEO) satellite net-
works with inter-satellite links (ISLs) are of particularly inter-
est among researchers for its advantages, such as the compa-
rable propagation delay with the terrestrial networks and the
relaxed power requirement for both the satellites and the hand-
held devices. Notwithstanding all its merits, the drawbacks
induced by the dynamics of the LEO satellite networks, e.g.
the variance of the ISL length and the discontinuous operation
of the ISLs, make the routing a daunting challenge. Differ-
ent mechanisms of the early work were proposed to shield
the dynamics of the constellations [1]. Yet due to the non-
homogeneous distribution of the users around the globe, the
satellites covering the dense population areas may easily get
congested while others are under-utilized in the meantime [2],
it is also pivotal for the routing algorithm to be able to achieve
a balanced traffic distribution in the LEO satellite networks.
This paper proposes the agent-based dynamic routing (ADR)
mechanism in the LEO satellite networks. There are two kinds
of agents, namely the roaming agents and the fixed agents. The
roaming agents traverse randomly across the satellite network
to collect the link status information of the satellites and at
the same time transfer the information of the past satellites
to the fixed agents of the satellites. On the other hand, the
fixed agents that locate at the satellites estimate the link cost
based on the link status information obtained from the roaming
agents and apply the HALO algorithm [3] proposed for the
load balancing of the terrestrial networks to update the routing
table of the satellites. The simulation results corroborate the
improved gains of the proposed mechanism.
This paper is organized as follows. Section II presents
the related works along with the motivation of the proposed
mechanism. Section III describes the network model briefly
while the detailed ADR mechanism is delineated in Section IV.
Section V gives the simulation results and discussion. Finally
concluding remarks are drawn in Section VI.
II. RELATED WORKS
Routing in the LEO satellite networks has been the subject
of considerable study since 1990s [1]. The routing mechanisms
originally proposed were connection-oriented [4]. As the IP
traffic gradually dominates the Internet, the connectionless
routing becomes the hot topic. The former studies mainly
focused on the mechanism to shield the dynamics of the
LEO satellite networks and find a feasible path to reduce the
end-to-end delay [5]. This type of routing algorithm assumes
the light traffic load on the LEO satellite networks, thus the
propagation delay is the main factor that contributes to the
whole end-to-end delay. As the earth is mostly covered by sea
and the majority of the hot spots are located in the northern
hemisphere, the routing with load-balancing mechanism is
of great importance to avoid congestion and improve the
utilization efficiency of the LEO satellite networks.
As has been pointed out in [6] that the first consideration
of the routing in the satellite network is the dissemination of
the network state information. From this point of view, Franck
et al. [7] firstly provided the criterion for the classification of
the routing algorithm, that is the static routing vs. the adaptive
routing which can be further classified into the isolated routing
vs. the non-isolated routing. Then, they evaluated a number
of routing strategies and pointed out that the adaptive and
non-isolated routing is superior provided that the signaling
overhead of the distribution of the network status is properly
managed.
Motivated by the traffic concentration at the higher lati-
tudes, Mohor
ˇ
ci
ˇ
c et al. [8] proposed the alternate link routing
(ALR) strategy. ALR tries to distribute the traffic via alter-
nate links, which effectively reduces the traffic peak. Then
Mohor
ˇ
ci
ˇ
c et al. [9] proposed the traffic dependent routing
(TCD) which classifies the traffic as different classes. TCD
tries to optimize the routes for each traffic class with specified
QoS requirement, e.g. routing the throughput-intensive traffic
via the path with the maximum available bandwidth. The
simulation model in both [8] and [9] assumes that the network
978-1-4673-7686-0/15/$31.00 ©2015 IEEE