1. Introduction
The autonomous underwater vehicle is currently the only equipment capable of deep-sea work, and it plays an important role in the detection of crashed aircraft and submarine optical cables [
1,
2]. AUVs that are unmanned and cableless work autonomously in the marine environment. Due to the complex marine environment and the different task requirements of AUVs, the problem of AUV control is one of the current research hotspots in this field [
3,
4].
AUV control mostly refers to trajectory tracking control, which tracks a predesigned desired trajectory and pursues higher tracking accuracy for the desired trajectory [
5,
6]. When AUVs perform tasks such as detection of crashed aircraft, the desired trajectory is generally a lawnmower or sawtooth pattern. Performing these tasks, the detection range of the sonar sensor carried by the AUV has overlapping regions. Therefore, in the process of tracking the desired trajectory, there is no need to pursue higher tracking accuracy, but the desired trajectory is changed from a curve to a cylinder with the curve as the centerline, ensuring that the actual trajectory can meet the detection requirements within this cylinder. That is, there is a change from curve tracking to region tracking [
7,
8,
9,
10]. The reason to do this is that the pursuit of higher tracking accuracy in curve tracking will cause frequent chattering of the control amount, resulting in increased energy consumption. The region tracking does not pursue higher tracking accuracy, but only requires the actual trajectory to be in the desired region, so as to reduce the frequent chattering of control quantity and reduce energy consumption.
At present, AUV region tracking control is mainly realized based on the scheme of boundary potential energy function. In [
7], an AUV region tracking control scheme based on the boundary potential energy function was first proposed. This scheme first constructs a boundary potential energy function, and then integrates the potential energy function into the Lyapunov function, so as to deduce the control law. Reference [
8] proposed several approximate Jacobian regions reaching control for robots with uncertain kinematics and dynamics. In this approach, the desired region can be specified as several inequalities. To solve the problem that hydrodynamic parameters of AUV are difficult to obtain accurately, reference [
9] proposed a regional tracking control method combining an RBF neural network and boundary potential function. Reference [
10] proposed an adaptive region tracking method for an AUV with a redundant thruster system, which carries out the region tracking of AUV in the presence of the thruster fault. Reference [
11] addressed the region tracking control problem for AUV with control input saturation, and proposed an adaptive region tracking control scheme based on nonlinear error transformation and prescribed performance control. In [
12], the problem of region tracking control based on region constraints was studied. Combined with the techniques of boundary potential energy function and dynamic surface, an adaptive region tracking controller was designed. In [
13], an adaptive region tracking control scheme with prescribed transient performance for an AUV with thruster fault was proposed, which can ensure that the tracking error is guaranteed to remain within the prescribed tracking performance.
Further research shows that there is a problem of overshoot in the AUV tracking of the desired region utilizing the above literature schemes, that is, the actual trajectory of the AUV will often exceed the outer boundary of the desired region. To solve this problem, reference [
14] designed a control compensation term based on the target speed. This scheme can reduce the number of overshoots to a certain extent, but it still cannot eliminate the overshoot. Reference [
15] deduced the control law by constructing a continuous and differentiable piecewise Lyapunov function instead of the boundary potential energy function, and then realized the region tracking through the compensation control scheme. This scheme relies on artificially reducing the outer boundary of the desired region in the controller to realize that the actual trajectory does not exceed the preset outer boundary of the desired region. The function of the outer boundary of the desired region is not fully exerted, which makes the scheme energy-intensive.
Based on the above analysis, aiming at the overshoot problem and high energy consumption of the previous AUV region tracking control schemes, this paper proposes a prediction-based region tracking control (PRTC) scheme for AUV. This scheme considers the large inertia characteristics of AUV, predicts the future position of AUV based on the position information of past time series, and compares it with the outer boundary of the desired region for predictive control. The digital simulation results, which are based on Odin AUV, show that this scheme can complete no overshoot region tracking control, and has obvious advantages in reducing energy consumption.
Then, based on the PRTC scheme, this paper further studies the relationship between the desired region range (the distance from the outer boundary of the desired region to the centerline) and the control output. It is found that in the process of controlling the movement of AUV from the outer boundary to the centerline, the controller output is determined by the error (the distance from AUV to the centerline). The larger the error, the larger the control output. When the desired region is large and the AUV is close to the outer boundary, far from the centerline, the control output is large, which will lead to control output saturation and increased energy consumption. Aiming at this problem, this paper proposes a control law optimization scheme considering the desired region range. The scheme takes the relative value of the distance between the AUV from the centerline and the desired region range as the error signal, and uses this signal as the input signal of the controller. The digital simulation experiment results based on ODIN AUV show that the control law optimization scheme can achieve non-saturated control output and is not affected by the desired region range. Compared with the PRTC scheme, the control law optimization scheme can further reduce energy consumption. The main contributions of this paper are as follows.
In this paper, A PRTC scheme for AUV is proposed, which can conquer the overshoot problem of the traditional region tracking scheme and has lower energy consumption than the traditional scheme.
A control law optimization scheme considering the range of the desired region is proposed, which improves the ability of the PRTC scheme to resist control output saturation and further reduces consumption.
The remainder of this paper is organized as follows.
Section 2 elaborates the proposed prediction-based region tracking control scheme in this paper, and describes the idea of the scheme and its implementation process.
Section 3 elaborates the proposed control law optimization, considering the desired region range scheme, and explains the idea and implementation process of this scheme. In
Section 4, the schemes proposed in
Section 2 and
Section 3 are verified by a digital simulation experiment. Finally, conclusions are drawn in
Section 5.
2. Prediction-Based Region Tracking Control Scheme
This section analyzes the reasons for the overshoot of the traditional region tracking scheme, explains the idea and implementation process of the PRTC scheme proposed in the paper, completes the design of the controller, and gives the stability proof of the controller. Simulation experiments are carried out to compare and verify the feasibility of the PRTC scheme.
2.1. Analyze the Causes of Overshoot in Traditional Schemes and the Idea of the PRTC Scheme
In order to explain how the PRTC scheme solves the problem of overshoot in the traditional scheme, this section analyzes the causes for the overshoot of the traditional scheme and expounds the idea of the PRTC scheme to solve this problem.
2.1.1. Analyze the Causes for the Overshoot of the Traditional Scheme
The traditional scheme [
7,
8,
9,
10,
11,
12] is to derive the control law by incorporating the boundary potential energy function into the Lyapunov function or directly constructing the piecewise Lyapunov function. When the AUV position does not exceed the outer boundary of the desired region, the Lyapunov function value is 0, and the controller does not adjust the AUV position at this time. Once the position of the AUV exceeds the outer boundary of the desired region, the Lyapunov function value is positive, and the controller adjusts the position of the AUV to make it move to the desired region. Based on the above control logic, it can be known that when the controller starts, the actual position of the AUV exceeds the outer boundary of the expected region, and overshoot occurs. At the same time, due to the large inertia of the AUV, the overshoot will continue for a period of time. It is concluded that the fundamental reason for overshoot in traditional schemes is that AUV inertia is not considered, and there is lag in position adjustment.
The root cause of the overshoot in the traditional scheme is that the controller starts to adjust the position when the AUV position exceeds the outer boundary of the desired region.
2.1.2. The Idea of the PRTC Scheme
The basic idea of the PRTC scheme is that considering the large inertia characteristics of AUV, based on the past time-series position information to predict the subsequent position relationship between the AUV and the outer boundary of the desired region, allows one to carry out predictive control. The specific instructions are as follows.
As shown in
Figure 1, one must calculate the speed
and acceleration
of the AUV moving towards the outer boundary of the desired region at time
through the positions
,
, and
at time
,
, and
. In this paper, considering the large inertia characteristics of AUV,
and
at time
are used to calculate the position
,…,
of the AUV at the subsequent time, so as to realize the prediction of the position of the AUV. When the AUV is in the desired region, the controller is not activated, and no force is given to the AUV in the direction of the centerline of the desired region, so the acceleration of the AUV moving in the direction of the outer boundary is less than zero. Therefore, after several control steps at time
, the AUV will move to the extreme position (the point farthest from the desired region center, as shown in
Figure 1), and after that, it will be reversed. When it is predicted that this extreme position exceeds the outer boundary of the desired region, the controller starts to pull the AUV toward the centerline of the desired region. In this way, in advance control at time
, the extreme position after
step will not exceed the outer boundary of the desired region.
According to the above scheme, the AUV can achieve no overshoot region control. However, the position of the AUV will fluctuate near the outer boundary of the desired region, and the random ocean current may cause the AUV to exceed the desired region. For this reason, the controller in the above-mentioned scheme will keep functioning after it is started, so that the AUV moves to the direction of the centerline of the desired region. After AUV returns to the centerline, the controller must be shut down. One must lay a good foundation for the follow-up region tracking.
Based on the above idea, the control flow of the PRTC scheme is shown in
Figure 2.
Figure 2 shows a complete control flow of the PRTC scheme. The controller has gone through a complete off-on-off process in a control flow. The specific process is as follows: (1)predict the extreme position of AUV; (2) determine whether the extreme position exceeds the outer boundary; (3) determine whether the controller has started based on the judgment result of the first step; (4) after the controller starts, judge whether the limit position extreme position exceeds the centerline; (5) based on the result of the previous step, the controller is shut down or kept in the current state; (6) when the controller is shut down, the control process ends.
2.2. The Implementation Process of the PRTC Scheme
This scheme is implemented in three steps. The relationship of parameters to be used is shown in
Figure 3, and the implementation process of each step is described in detail below with reference to
Figure 3.
Step 1: Calculate the speed and acceleration of the AUV moving towards the outer boundary of the desired region.
The speed
and acceleration
of the AUV moving towards the outer boundary of the desired region are obtained by using the distances of the AUV from the outer boundary of the desired region at times
,
, and
(
is the control step). Let the distances of the AUV from the centerline of the desired region at time
,
, and
be
,
, and
, respectively. The distance from the outer boundary of the desired region to the centerline is
. Then, the
and
can be expressed as:
Step 2: Predict the extreme position that the AUV can reach after time .
AUV has the characteristic of large inertia. Without additional control, it can be approximated that the movement trend of AUV after time
is the same as that at time
, that is, the acceleration after time
is replaced by
. Therefore, this paper uses Equations (1) and (2) to predict the AUV position after time
. Based on the displacement equation, the predicted position of the AUV for each step after time
is:
where the
represents the number of control steps.
When the AUV is in the desired region, the controller is not activated, which means that no force is given to the AUV in the direction of the centerline of the desired region, and the acceleration of the AUV moving in the direction of the outer boundary is less than zero. Therefore, after reaching the extreme position, the AUV will move to the centerline of the desired region, as shown in
Figure 1. This extreme position
can be found by:
where the
,
,
are the distances from the AUV to the centerline of the desired region at
,
,
, respectively. It is worth noting that in the above position prediction process based on Equations (3) and (4), parameter
is an intermediate parameter, which does not need to be expressed concretely.
Step 3: Startup and Shutdown of the Controller.
The controller will start when the extreme position is predicted to exceed the outer boundary. In order to convert the above process into a logical signal, design the startup functions as:
As shown in
Figure 3, if
, this indicates that the AUV extreme position exceeds the outer boundary of the desired region; then, for
in Equation (5), the controller is started. Otherwise, do not operate the controller.
The above controller remains on after startup until the AUV is pulled back to the centerline of the desired region, then the controller is shut down. In this way, the AUV will be in a position with a large range of free movement in any direction in the desired region, thereby improving the ability to resist ocean current disturbance and laying a good foundation for subsequent region tracking.
By comparing
, judge whether AUV has reached the centerline of the desired region. In this paper, the
above the centerline is positive, and below the centerline is negative. Transform the judgement result into a logical signal, and design the shutdown functions as:
In Equation (6), if , the controller is shut down, otherwise, the controller does not operate.
The overall process of startup and shutdown in the controller is shown in
Figure 2. In order to realize the overall process, it is necessary to design the overall work functions based on the controller startup and shutdown functions. The work functions are as follows:
where
is the work functions of the controller in this paper, and
and
respectively represent the working state of the work functions at time
and time
. The
is a constraint condition, indicating that
and
cannot be 1 at the meantime.
By Equation (7), the principle of the controller’s working functions is explained. If
, then
, and the controller is started. If
, there are two cases: in the first case,
,
is the same as
, which means that the controller continues to use the working state of the previous step. In the second case,
, then
, and the controller is shut down. It can be seen that the proposed work functions can realize the overall work process as shown in
Figure 2.
In the real applications of the PRTC scheme, the first three control steps are only used to calculate the
,
, and
. Until the fourth step, according to the flow in
Figure 2, the PRTC scheme is used for AUV region tracking control.
2.3. Design Controller of PRTC Scheme
This section completes the controller design based on the implementation process of the PRTC scheme described in
Section 2.2. Firstly, the dynamic model in this paper is described, and then the design idea and the implementation of the PRTC controller are given.
2.3.1. Dynamic Model
ODIN AUV is a typical AUV. It has an open dynamic model. Many references use this model to verify the effectiveness of the method [
13,
16]. Therefore, the controller design and simulation in this paper also use the ODIN AUV model. The ODIN AUV dynamic model is expressed as:
where,
,
,
,
,
; the
is a vector about AUV’s position and attitude in the earth-fixed frame.
is the transformation matrix;
is the inertia and added mass matrix;
and
are rigid-body and Coriolis matrixes;
is the hydrodynamic damping matrix;
is a vector of gravity and buoyancy forces and moments;
is the force/moment applied to the AUV. The specific values of each parameter in ODIN AUV can be found in [
15,
16].
2.3.2. Design Idea and the Implementation of Controller of PRTC Scheme
The control law of the traditional region tracking scheme is obtained based on the method of converging the boundary potential energy functions or the piecewise Lyapunov function to 0. This control law cannot adjust the position of the AUVs in the desired region. In addition, the control law aims to converge the AUVs to the outer boundary of the desired region, so it can only realize the process of pulling the AUVs outside the desired region to the boundary of that. Different from the traditional scheme, the PRTC scheme requires the control law to be able to adjust the position of AUV in the desired region, and at the same time requires the control law to be aimed at the convergence of AUV to the centerline. Hence, the controller law of PRTC scheme cannot be obtained by the traditional scheme.
To meet the above requirements, this paper is inspired by the event triggered control scheme [
17]. The event trigger control’s output is adjusted non-periodically by discrete trigger conditions. Based on the above discussion, the controller of PRTC adopts the controller working functions obtained in
Section 2.2 as the discrete control trigger condition to switch between interrupted control output and continuous control output. So, the control law of the PRTC scheme is expressed as:
where, the equation in { } is the typical sliding mode control law, and the derivation process of the control law is common;
;
is the sign function;
,
are the control parameters.
It can be seen from Equation (7), if , then the control law is , indicating that the controller has no control output, that is, the controller is shut down; if , the control is , representing that the controller has started and the control output has been adjusted to follow typical sliding mode control law.
2.4. Analysis of Controller Stability of PRTC Scheme
To prove that the AUV can converge to the centerline of the desired region based on the control law in Equation (9), controller stability analysis based on Lyapunov theory is required.
Consider the following Lyapunov function as:
Differentiating both sides of Equation (10) with respect to time, one has:
where
where, the
is a small positive number.
According to
Figure 3 and the definition of the dynamic model,
is equivalent to
. Substituting Equations (8) and (12) into Equation (11), one has:
The control law in Equation (7) will present two control states based on the different values of . Firstly, one must analyze the situation of .
Taking
in Equation (7) and bringing it into Equation (13), one has:
where,
represents the Euclidean norm of the vector.
According to the Lyapunov stability theory, the controller can converge to 0 from any initial value at , that is, the AUV can converge to the centerline of the desired region from any position.
For
, the error variable will diverge because the controller does not have any control output. However, based on the relationship between
and the outer boundary of the desired region in
Section 2.2, it can be seen that
is bounded by
. This is due to the actual assumption that the
is bounded, i.e.,
for positive bound
. So, the divergence of the error variable
is bounded. As
increases, the
jumps from 0 to 1, and the divergence of
in
stage will be the initial value of
stage. According to the above,
stage does not affect the stability of the controller.
The feasibility of the prediction-based AUV region tracking control scheme will be verified in
Section 4 by simulation experiments with the traditional scheme based on boundary potential function.
5. Conclusions
In this paper, the problem of AUV region tracking is studied. Aiming at the problem of overshoot in traditional region tracking schemes, a prediction-based AUV region tracking (PRTC) scheme is proposed. The PRTC scheme realizes prediction control based on the relative motion relationship between AUV and the outer boundary of the desired region. The digital simulation results show that the PRTC scheme can realize the region tracking without overshoot. The digital simulation results show that compared with the traditional scheme, the energy consumption of the PRTC scheme in the three directions of sway, surge, and heave is reduced by 42.0%, 14.5%, and 3.0%, respectively, and the total energy consumption is reduced by 32.3%. In addition, because the PRTC scheme is different from the traditional scheme, it adds the control process to make the AUV return to the center of the desired region, resulting in the control output of the PRTC scheme being affected by the range of the desired region. As the range of the desired region increases, the control output saturation becomes more serious. To solve this problem, this paper proposes a control law optimization scheme considering the desired region. By transforming the error variables in the original scheme, the influence of the desired region on the control output of PRTC is eliminated. Theoretical analysis and simulation results show that the control output of the optimized control law is not affected by the desired region, and the energy consumption is reduced by 3.7% compared with that before control law optimization.
Deficiency of proposed method and further work plan: (1) the supplementary simulation in this paper shows that, in the presence of uncertainty, although the performance of the proposed scheme in the paper in terms of energy consumption is still better than that of the traditional scheme, the reduction of energy consumption is less than that without uncertainty. Therefore, in the next step, the authors will study how to achieve the goal of keeping low energy consumption while tracking the desired region in the presence of uncertainty. (2) In addition, the performance of the proposed scheme in flow perturbations and rapid changes of the medium, say pycnocline, also needs to be further studied.