1 INTRODUCTION
Path planning in dynamic environment is one of the most
challenging issues for autonomous vehicle. Traditionally, path
planning problem is defined as: given a vehicle A and a target G
that are moving, planning a trajectory enables the vehicle to catch
the target and satisfies some specified constrains while avoiding
obstacle O. Also, each of the obstacles can be either mobile or
immobile in the environment during the planning procedure. All
the waypoints of the vehicle should be available so as the
constraints of the vehicle, including the kinematics and dynamics
constraints would be satisfied. Besides, the path needs to consider
some optimal criterion, such as the shortest travelled distance, the
minimum consumed time or energy. The traditional methods in
this field have achieved some applications in static environments
[1-2]. However, the robot has to plan in real-time once it keeps in
dynamic environment where the target and the obstacles are
moving. This poses a great challenge when planning in unknown
dynamic environment.
In order to solve this problem, many methods have been
proposed in past decades. One of these methods is mixed integer
linear programming (MILP). It calculates the trajectory points
with collision-avoiding constraints [3-4]. By introducing binary
variables and linearization technique, MILP computes the actions
that will be taken by vehicle in the next step. However, MILP only
supplies the next step actions instead of multiple steps. In order to
improve the planning performance, receding horizon control
(RHC) has been used in path planning [5-12]. After that, some
linear constraints in future time can be added in MILP [7] and then
a solution sequence can be derived. This method shows the
feasibility in static environment [8].
An important condition in the above MILP method is the
linearization of all the constraints which are built on many
assumptions [13]. However, the kinematic and dynamic of a real
This work is supported by National Nature Science Foundation under
Grant 61203331, Hubei Province Key Laboratory of Systems Science in
Metallurgical Process (Wuhan University of Science and Technology)
(No.Z201301), and Henan Provincial Open Foundation of Control
Engineering Key Lab of China (No.KG2011-01).
robot are definitely nonlinear. So, the MILP solution is ready to be
much conservative even no solution. To solve this problem, this
paper focuses on the relaxation of the linear constraints by
introducing the quadratic constrained quadratic programming
(QCQP). In this new method, it is unnecessary to linearize the
kinematic or the dynamic function but reserve the original
quadratic cost functions. Finally, a QCQP model is built on the
receding horizon control framework. This method has potential
application in variable control, such as autonomous vehicle
control [14].
This paper is arranged as follows. Section II introduces some
assumptions and the state equations of the robot. The QCQP
model is established in Section III which mainly investigates how
to acquire the constraints in the receding horizon framework.
Some simulations are demonstrated in Section IV. A brief
conclusion is drawn in Section V.
2 STATE EQUATIONS OF THE MOBILE
ROBOT
2.1 Some Assumptions and Nomenclatures
This method will plan the accelerations of the mobile robot,
a=(a
x
, a
y
), which is supposed moving on a 2-D plane. We
assume that the current position and velocity of the robot
are known by means of sensors. In the following
expressions, the subscript A denotes the robot, O the
obstacles, and G the target. The control period is described
by
τ
. Some variables are defined as the following.
i) Vehicle velocity: V
A
= (v
Ax
, v
Ay
), V
A
= |V
A
|;
Vehicle position: L
A
=(l
Ax
, l
Ay
).
ii) Obstacle velocity: V
Oi
= (v
Oix
, v
Oiy
), V
Oi
= |V
Oi
|;
Obstacle position: L
Oi
=(l
Oix
, l
Oiy
); threat radius of
the obstacle: r
Oi
, where i denotes subscript of the
obstacle.
iii) Target velocity: V
G
= (v
Gx
, v
Gy
), V
G
= |V
G
|; Target
position: L
G
=(l
Gx
, l
Gy
,).
Receding Horizon Control for Mobile Robot Path Planning in Unknown Dynamic
Environments
Yang Chen
1
, Lei Cheng
1
, Huaiyu Wu
1
, and Yanhua Yang
2
1. Wuhan University of Science and Technology, Wuhan 430081, China
E-mail: chenyag@wust.edu.cn
2. Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China.
E-mail: yangyh15@gmail.com
Abstract: Path planning is one of the critical issues in mobile robot applications. Traditional methods for path planning
in unknown dynamic environment generally plan one step rather than multiple controlling steps. This paper proposes an
approach with multiple controlling steps which integrates receding horizon control (RHC) for mobile robot path planning
in which the obstacle avoidance problem is converted into mathematic constraint. Besides, the distance from target to
robot is optimized through involving a quadratic cost function. Finally, a quadratic constrained quadratic programming
(QCQP) is generated, from which a sequence including multiple control steps is easily obtained. Based on the proposed
approach, simulations are demonstrated to verify the effectiveness of the novel method.
Key Words: Path planning, Receding Horizon Control, Constraints
1505
978-1-4799-3708-0/14/$31.00
c
2014 IEEE