Cooperative Area Reconnaissance for Multi-UAV in
Dynamic Environment
Jie Chen
∗ †
, Wenzhong Zha
∗ †
, Zhihong Peng
∗ †
, Jian Zhang
∗ †
∗
School of Automation, Beijing Institute of Technology, Beijing, China 100081
†
Key Laboratory of Complex System Intelligent Control and Decision of Ministry of Education
Beijing Institute of Technology, Beijing, China 100081
Email: chenjie@bit.edu.cn, wenzhong@bit.edu.cn, peng@bit.edu.cn, zhangjian916@gmail.com
Abstract—The increasingly complex battlefield environment
has put forward higher requirement on Unmanned Aerial Vehicle
(UAV) system, where the cooperative area reconnaissance (CAR)
is the primary task for multi-UAV. However, the current research
results hardly balance the optimality and real-time property.
Especially, how to avoid and process emergent threats is rarely
considered for UAV formation. So this paper researched on the
problems of CAR for multi-UAV in dynamic environment to
obtain optimum efficiency on the premise of ensuring real-time.
Firstly, the mathematical model and optimization framework
were established. Then the idea of Model Predictive Control
(MPC) was introduced to process this model and an improved
Particle Swarm Optimization (PSO) algorithm based on Sim-
ulated Annealing (SA) was proposed to solve the optimization
problem. Furthermore, the termination condition of searching
was defined and processing strategies in multiple emergent condi-
tions were represented specially. Finally, analysis and comparison
of the results from established simulation platform verified that
the methods proposed in this paper could control the UAVs
avoiding the static and mobile threats effectively, accomplishing
task perfectly with more than 90% reconnaissance coverage rate
and the run-time of each prediction step was only 1.3892s.
Index Terms—Multi-UAV, Cooperative Area Reconnaissance,
Dynamic Environment, MPC-SAPSO, Real-time
I. INTRODUCTION
The complex and dynamic battlefield environment will
make cooperative combat by multi-UAV become the primary
air combat mode in the future. And the principle requirement
for multi-UAV systems is the ability to carry out cooper-
ative area reconnaissance (CAR), because this is the most
direct route to obtain the battlefield intelligence and situation
roundly. After that, the operational security of UAVs can be
ensured and the enemy targets will be struck in favorable
opportunity. Finally, the task period of cooperative combat
could be shortened to improve combat efficiency.
The CAR for multi-UAV is a typical problem of multi-
agent coordinated control. That is, it needs to implement
effective control for multiple UAVs to satisfy specific task
requirements and constraint conditions to realize coverage
reconnaissance and target searching in unknown battlefield
environment with highest quality and least price. This problem
refers to coordinated control of multiple vehicles, model of
region search and online routes planning. About coordination
control, the hierarchical and distributed framework [1] is
introduced usually. When concerning the model of region
search, the discrete search graph [2] is used mostly to describe
the battlefield environment. With regard to online routes plan-
ning, the general methods include mathematical programming
method [3] and intelligent optimization algorithm (such as
genetic algorithm [4], ant colony algorithm [5], evolutionary
algorithm [6] and particle swarm optimization algorithm [7]).
As the intelligent optimization algorithm has the capability of
intelligent searching and the particular advantages on solving
online routes planning problem, it has been a hot point in
studying.
However, from the existing related literatures, we find that
the battlefield environment is very complex and changeful,
while current description of environment information is only
based on great simplification and has little consideration on
dynamic situation especially the emergent threats. On the
online routes planning of multi-UAV, most researches pursue
the optimum of solution while neglect the requirement of real-
time property.
Aim at the above defects, this paper endeavors to research
on CAR for multi-UAV in dynamic environment to obtain
optimum efficiency on the premise of ensuring real-time.
Firstly, the mathematical model and optimization framework
will be established based on deep analysis on CAR problems.
Then the idea of MPC will be introduced to process this model
and an improved PSO algorithm (SAPSO) will be proposed
to solve the optimization problem. Finally, the multi-thread
technology will be applied to set up a simulation platform to
verify the feasibility and validity of the methods proposed in
this paper.
II. MODELING ON THE CAR PROBLEMS FOR MULTI-UAV
A typical task scenario of CAR for multi-UAV can be
described as follow: There are many potential targets and
threats in enemy area and they may be partially known or
completely unknown. Multiple heterogeneous UAVs fly into
this area to reconnoiter and search worthy information using
airborne detection equipments (e.g. camera, infrared detector,
or laser detector). In some particular cases, such as carrying
weapons on UAVs, they can attack the enemy targets based
on this reconnoitered information after getting confirmation
and authorization of commander. The objective of CAR is
to ensure the UAV formation obtains maximum benefits (the
level of knowledge about battlefield environment) and pays
978-1-4673-5769-2/13/$31.00 ©2013 IEEE