
IEEE/CAA Journal of Automatica Sinica
Citation: | S. Gao, Z. Peng, H. Wang, L. Liu, and D. Wang, “Long duration coverage control of multiple robotic surface vehicles under battery energy constraints,” IEEE/CAA J. Autom. Sinica, vol. 11, no. 7, pp. 1695–1698, Jul. 2024. doi: 10.1109/JAS.2023.123438 |
Dear Editor,
This letter addresses long duration coverage problem of multiple robotic surface vehicles (RSVs) subject to battery energy constraints, in addition to uncertainties and disturbances. An anti-disturbance energy-aware control method is proposed for performing coverage task of RSVs. Firstly, a centroidal Voronoi tessellation (CVT) is used to optimize the partition of the given coverage area. The optimal position for each vehicle corresponds to the centroid of the Voronoi cell. Secondly, by consisting two battery energy control barrier functions, an energy-aware kinematic guidance law is designed to drive each RSV to the optimal position. Finally, an anti-disturbance fixed-time kinetic control law is designed for each RSV to track the desired speed based on the fixed-time extended state observer and nonlinear tracking differentiator. By the control method, RSVs are capable of achieving the long duration coverage within the given coverage area. Simulation results verify the effectiveness of the proposed anti-disturbance energy-aware control method for multi-RSV.
There is a surge of interest in motion control of RSVs because they are vital tools in exploring the oceans [1]–[3]. Cooperative control enables RSVs to perform more challenging task over a single RSV in terms of enhanced efficiency and effectiveness [4], [5]. In particular, coverage control is a typical cooperative control of RSVs. Such motion control scenario can find numerous missions in practice such as maritime patrol and environment monitoring [1], [6]–[9]. Coverage control is proposed for unmanned ground vehicles [6], unmanned aerial vehicles [7], and robotic surface vehicles [8]. However, the aforementioned works [6]–[9] do not explicitly guarantee the persistent coverage performance. In practice, the vehicles may not execute task persistently because of the battery energy constraints.
Motivated by the above observations, this letter aims to design an anti-disturbance energy-aware control method for RSVs to execute long duration coverage task. The main contributions of this letter can be summarized as: 1) In contrast to the works in [4], [5] where the objectives are to maintain formation, the proposed control method for multi-RSV to achieve coverage task within a given coverage area. 2) In contrast to the existing works [6]–[9] where the persistent coverage performance may not be guaranteed definitely, an energy-aware control method based on two battery energy control barrier functions (BE-CBFs) are designed to achieve long duration coverage mission. 3) In contrast to the existing coverage control methods in [8], [9] where the convergence time of estimation errors can not be calculated in a prescribed time, a fixed time extended state observer (FTESO) is proposed such that the estimation errors are convergent to zero within a fixed time.
Problem formulation: Consider the dynamics of RSVs as
{˙xi=uicos(ψbi)−υisin(ψbi)˙yi=uisin(ψbi)+υicos(ψbi)˙ψbi=rimui˙ui=fui(ui,υi,ri)+τui+τwuimυi˙υi=fυi(ui,υi,ri)+τwυimri˙ri=fri(ui,υi,ri)+τri+τwri | (1) |
where xi, yi, and ψbi are the position and orientation in an earth-fixed inertial frame; ui, vi, and ri are the velocities in the surge, sway, and yaw directions respectively in a body-fixed frame; mui, mυi, and mri are the mass of RSVs; fui(⋅), fυi(⋅), and fri(⋅) represent unknown nonlinear functions; τui and τri denote the surge force and yaw moment; τwui, τwυi, and τwri denote the time-varying environmental disturbances. Denote Ui=√u2i+υ2i as the total speed of the ith RSV, and the dynamics (1) of the RSV can be rewritten as
{˙xi=Uicos(ψi),˙yi=Uisin(ψi),˙ψi=ri+˙βimui˙Ui=cos(βi)(fui+τwui)−2sin2(βi2)τui+sin(βi)(fυi+τwυi)mui/mυi+τuimri˙ri=fri+τri+τwri | (2) |
where ψi=ψbi+βi is the course angle with βi=atan2(υi,ui).
The charging and discharging model of the battery is stated as ˙Ei=−δKch+(1−δ)Ken, where Ei is the battery energy level of the ith RSV; Kch>0/Ken>0 represents the case of the worst/fastest battery discharging/charging; δ∈[0,1] is the switching coefficient of battery charging and discharging.
Design and analysis: Step 1: The first step is optimizing the partition of the given coverage area. CVT is a powerful tool for the area partition. The centroid of each Voronoi cell can be defined as the optimal position for each RSV to track. Consider that N RSVs are randomly distributed in a closed and connected area G∈R2 that has to be covered. Associate a density function ϕ(g):g∈R+ with each point g∈G. Define locational cost function of the multi-RSV as H(P)=∑Ni=1∫Vi‖pi−g‖2ϕ(g)dg, where pi=[xi,yi]T is the position of ith RSV; P=[p1,⋯,pN] is position set of RSVs; ‖pi−g‖ denotes the Euclidean distance metric. Next, the Voronoi cell of the ith RSV can be defined as Vi={g∈G|‖g−pi‖≤‖g−pj‖,∀i≠j}. Then, it follows from [8] that the optimal position pci corresponds to the centroid of the Voronoi cell in a CVT is:
pci=∫Vigϕ(g)dg/MVi | (3) |
where MVi is the mass of the Vi with MVi=∫Viϕ(g)dg.
Step 2: Define the target tracking error as epi=pi−pci. Take the derivative of epi along (2), and one has ˙epi=qi−˙pci, where qi=[qxi,qyi]T=Ui[cos(ψbi),sin(ψbi)]T. To stabilize epi, an optimal total speed guidance law is designed as qci=−Kqiepi/√‖epi‖2+Δ2qi+˙pci, where Kqi∈R2 is the guidance law parameter vector and Δqi is a positive constant. Define the velocity tracking error as eqi=qi−qci and the time derivative of the epi can be rewritten
˙epi=−Kqiepi/√‖epi‖2+Δ2qi+eqi. | (4) |
To implement the long duration coverage control task, the position of charging station pbi should satisfy Echarge={εi(pi)∈R2|‖pi−pbi‖≤dch}, where εi(pi) is the minimum energy required for the ith RSV to return to the charging station and dch is the radius of the charging station. In order to maintain a desired energy reserve, the following inequalities Ei−Emin and E_{\max } - E_i>0 hold, where E_{\min} denotes the minimum capacity of battery storage and E_{\max} is the maximum capacity of battery storage. It means that each RSV can return to the charging station before limited battery energy depletion and leave the charging station before overcharging. According to the definition of the CBF, the functions h_{en} and h_{cn} are defined as h_{en}(z_i) = {E_i}-E_{\min}-{\varepsilon_i}(p_i) and h_{ch}(z_i) = E_{\max}\delta(s_{i0}) - E_i , where z_i = [p_i^T,E_i]^T , s_{i0} = \|p_i - p_{i0}\| , and \delta(s_{i0}) is a step function. If s_{i0} > d_{{ch}} , \delta(s_{i0}) = 1 ; otherwise, \delta(s_{i0}) = 0 . The safe set {\cal{C}}_{zi} in this letter is defined as {{\cal{C}}_{zi}} = \{{z_i}\left|{h_{en}(z_i)\geq 0},{h_{ch}(z_i)\geq 0}\right.\} .
It follows from the safe set that the functions h_{en} and h_{cn} can be described by an equivalent h(z_i) as h(z_i) = \text{min}\{ h_{en}(z_i),h_{ch}(z_i)\} . Thus, the speed guidance law q_{ci}^* = [q_{xi}^*,q_{yi}^*]^T can be obtained by the following quadratic programming problem:
\begin{split} &\arg\mathop{ \min }\limits_{{q_{ci}^*} \in {{\mathbb{R}}^2}} J_{i} (q_{ci}^*) = {\|{q_i- q_{ci}^*}\|^2}\\ & \qquad\text{s.t.}\; \;\dot h(z_i) \geq-{\alpha}h(z_i) \end{split} | (5) |
where \dot h(z_i) = (\frac{\partial h}{\partial p_i})^Tq_{ci}^*-\delta K_{ch}+(1-\delta)K_{en} . The optimal total speed of the ith RSV is derived as U_{ci} = q_{xi}^*\cos(\psi_{i})+q_{yi}^*\sin(\psi_{i}) .
Step 3: Define \psi_{ci} = \text{atan2}(q_{yi}^*, q_{xi}^*) and e_{\psi i} = \psi_{i}-\psi_{ci} , and the dynamics of e_{\psi i} is \dot{e}_{\psi i} = r_i+\dot{\beta}_{i}-\dot{\psi}_{ci} . To stabilize e_{\psi i} , a yaw guidance law is designed as
\begin{align} {r}_{ci} = -{k_{ri}e_{\psi i}}/{\sqrt{|e_{\psi i}|^2+\Delta_{ri}^2}}-\dot{\beta}_{i}+\dot{\psi}_{ci} \end{align} | (6) |
where k_{ri}\in {\mathbb{R}} is the guidance law parameter and \Delta_{ri} is a positive constant. Define {e_{ri}} = {r_i} - r_{ci} , and the dynamics of the yaw target tracking error is rewritten as
\begin{align} \dot{e}_{\psi i} = -{k_{ri}e_{\psi i}}/{\sqrt{|e_{\psi i}|^2+\Delta_{ri}^2}}+{e}_{ri}. \end{align} | (7) |
Step 4: Due to the fact that the uncertainties and disturbances can affect the performance of the control system seriously [10]–[12], it is necessary to recover the uncertainties and disturbances. First, to obtain a smoother motion profile for RSV, let {U}_{ci} and { r}_{ci} pass through two nonlinear tracking differentiators as
\begin{align} \left\{ {\begin{aligned} &{\dot {\hat{U}}}_{ci} = { U}_{di}-k_{i1}^U\text{sig}^{1-1/k_{i3}^U}(\hat{ U}_{ci}-{ U}_{ci}) \\ &{\dot U}_{di} = -k_{i2}^U\text{sig}^{1-2/k_{i3}^U}(\hat{ U}_{ci}-{U}_{ci}) \\ &{\dot {\hat{r}}}_{i} = { r}_{di}-k_{i1}^r\text{sig}^{1-1/k_{i3}^r}(\hat{ r}_{i}-{ r}_{ci}) \\ &{\dot r}_{di} = -k_{i2}^r\text{sig}^{1-2/k_{i3}^r}(\hat{ r}_{i}-{r}_{ci}) \end{aligned} } \right. \end{align} | (8) |
where \hat{U}_{ci} is the estimate of the {U}_{ci} ; \hat{r}_{i} is the estimate of the {r}_{ci} ; {U}_{di} is the estimate of the \dot{U}_{ci} , and {r}_{di} is the estimate of the \dot{r}_{ci} . k_{i1}^U\in{\mathbb{R}}^+ , k_{i2}^U\in{\mathbb{R}}^+ , k_{i3}^U>2 , k_{i1}^r\in{\mathbb{R}}^+ , k_{i2}^r\in{\mathbb{R}}^+ , and k_{i3}^r>2 are the design parameters. Define e_{Ui}^c = \hat{U}_{i}-{U}_{ci} , e_{ri}^c = \hat{r}_{i}-{r}_{ci} , e_{Ui}^d = {U}_{di}-\dot{U}_{ci} , and e_{ri}^d = {r}_{di}- \dot{r}_{ci} . It follows from (8) that the dynamics of the e_{Ui}^c , e_{ri}^c , e_{Ui}^d , and e_{ri}^d are obtained as:
\begin{align} \left\{ {\begin{aligned} &{\dot e}_{Ui}^c = { e}_{Ui}^d-k_{i1}^U\text{sig}^{1-1/k_{i3}^U}{ e}_{Ui}^c \\ &{\dot e}_{Ui}^d = -k_{i2}^U\text{sig}^{1-2/k_{i3}^U}{ e}_{Ui}^c{U}_{di}-\ddot{U}_{ci} \\& {\dot e}_{ri}^c = { e}_{ri}^d-k_{i1}^r\text{sig}^{1-1/k_{i3}^r}e_{ri}^c \\& {\dot e}_{ri}^d = -k_{i2}^r\text{sig}^{1-2/k_{i3}^r}e_{ri}^c{r}_{di}-\ddot{r}_{ci}. \end{aligned} } \right. \end{align} | (9) |
From (2), the kinetics of the ith RSV can be rewritten as \dot{U}_i = g_{ui}+m_{ui}^{-1}\tau_{ui} and \dot{r}_i = g_{ri}+m_{ri}^{-1}\tau_{ri} , where g_{ui} = m_{ui}^{-1}(\cos ({\beta_{i}})(f_{ui}+ \tau_{wui})- 2\sin^2(\frac{\beta_i}{2})\tau_{ui}+\sin ({\beta_{i}})(f_{\upsilon i}+\tau_{w\upsilon i})m_{ui}/m_{\upsilon i}) and g_{ri} = m_{ri}^{-1}(f_{ri}+ \tau_{wri}). Two fixed-time ESOs are designed to estimate the unknown functions g_{ui} and g_{ri}
\begin{split}& {\dot{\hat{U}}}_i = -k_{i1}^u {\lceil{\tilde U}_i\rfloor}^{\alpha_{ui}}-k_{i2}^u{\lceil{\tilde U}_i\rfloor}^{\beta_{ui}} + {{\hat g}_{ui}} + m_{ui}^{-1}{\tau_{ui}} \\ &{\dot{\hat{g}}}_{ui} = -k_{i3}^u {\lceil{\tilde U}_i\rfloor}^{2\alpha_{ui}-1}-k_{i4}^u{\lceil{\tilde U}_i\rfloor}^{2\beta_{ui}-1} + k_{i0}^u\text{sign}({{\tilde U}_i})\\& {\dot{\hat{r}}}_i= -k_{i1}^r {\lceil{\tilde r}_i\rfloor}^{\alpha_{ri}}-k_{i2}^r{\lceil{\tilde r}_i\rfloor}^{\beta_{ri}} + {{\hat g}_{ri}}+m_{ri}^{-1}{\tau_{ri}} \\ &{\dot{\hat{g}}}_{ri} = -k_{i3}^r {\lceil{\tilde r}_i\rfloor}^{2\alpha_{ri}-1}-k_{i4}^r{\lceil{\tilde r}_i\rfloor}^{2\beta_{ri}-1} + k_{i0}^r\text{sign}({{\tilde r}_i}) \end{split} | (10) |
where k_{ia}^u\in {{\mathbb{R}}^+} and k_{ia}^r\in {{\mathbb{R}}^+}(a = 1,2,3,4) are the observer gains; \hat {U}_i , \hat {r}_i , \hat g_{ui} , and \hat g_{ri} are the estimates of U_i , r_i , g_{ui} , and g_{ri} respectively; {\tilde U_i} = {\hat U_i} - {U_i} , {\tilde r_i} = {\hat r_i} - {r_i} , \alpha_{ui} \in (1-\varepsilon_{ui},1) , \beta_{ui} = 1/\alpha_{ui} , \alpha_{ri} \in (1- \varepsilon_{ri},1), and \beta_{ri} = 1/\alpha_{ri} with \varepsilon_{ui} and \varepsilon_{ri} being the small constants; k_{i0}^u>0 and k_{i0}^r >0 .
Assumption 1: For unknown functions g_{ui} and g_{ri} , there are g_{ui}^*\in {\mathbb{R}}^+ and g_{ri}^*\in {\mathbb{R}}^+ such that |\dot g_{ui}|\le g_{ui}^* and |\dot g_{ri}|\le g_{ri}^* .
Step 5: Let {\tilde g_{ui}} = {\hat g_{ui}}-{g_{ui}} and {\tilde g_{ri}} = {\hat g_{ri}}-{g_{ri}} . It follows from equation (10) that the error dynamics can be expressed as:
\begin{split} &{\dot{\tilde{U}}}_i = -k_{i1}^u {\lceil{\tilde U}_i\rfloor}^{\alpha_{ui}}-k_{i2}^u{\lceil{\tilde U}_i\rfloor}^{\beta_{ui}} + {{\tilde g}_{ui}} \\& {\dot{\tilde{g}}}_{ui}= -k_{i3}^u {\lceil{\tilde U}_i\rfloor}^{2\alpha_{ui}-1}-k_{i4}^u{\lceil{\tilde U}_i\rfloor}^{2\beta_{ui}-1} - {\dot g}_{ui} + k_{i0}^u\text{sign}({{\tilde U}_i})\\ & {\dot{\tilde{r}}}_i = -k_{i1}^r {\lceil{\tilde r}_i\rfloor}^{\alpha_{ri}}-k_{i2}^r{\lceil{\tilde r}_i\rfloor}^{\beta_{ri}} + {{\tilde g}_{ri}} \\& {\dot{\tilde{g}}}_{ri} = -k_{i3}^r {\lceil{\tilde r}_i\rfloor}^{2\alpha_{ri}-1}-k_{i4}^r{\lceil{\tilde r}_i\rfloor}^{2\beta_{ri}-1} - {\dot g}_{ri} + k_{i0}^r\text{sign}({{\tilde r}_i}). \end{split} | (11) |
Define the tracking errors as e_{Ui} = U_i-\hat{U}_{ci} and e_{ri} = r_i-\hat{r}_{ci} . Taking the derivatives of e_{Ui} and e_{ri} yields
\begin{align} \left\{ \begin{aligned} &\dot{e}_{Ui} = g_{ui}+m_{ui}^{-1}\tau_{ui}-\dot{\hat{U}}_{ci}\\ &\dot{e}_{ri} = g_{ri}+m_{ri}^{-1}\tau_{ri}-\dot{\hat{r}}_{ci}. \end{aligned} \right. \end{align} | (12) |
Two anti-disturbance fixed-time kinetic control laws are designed
\begin{align} \left\{ {\begin{aligned} &\tau_{ui} = m_{ui}(-k_{i1}^{\tau}{\lceil{e}_{Ui}\rfloor}^{\alpha_{ui}^{\tau}}-k_{i2}^{\tau}{\lceil{e}_{Ui}\rfloor}^{\beta_{ui}^{\tau}}- {{\hat g}_{ui}}+\dot{\hat{U}}_{ci}) \\ &\tau_{ri} = m_{ri}(-k_{i3}^{\tau}{\lceil{e}_{ri}\rfloor}^{\alpha_{ri}^{\tau}}-k_{i4}^{\tau}{\lceil{e}_{ri}\rfloor}^{\beta_{ri}^{\tau}}- {{\hat g}_{ri}}+\dot{\hat{r}}_{ci}) \end{aligned} } \right. \end{align} | (13) |
where k_{ia}^{\tau}\in {{\mathbb{R}}^+} (a = 1,2,3,4) are the control gains. Then, the tracking error systems (12) can be put into \dot{e}_{Ui} = -k_{i1}^{\tau}{\lceil{e}_{Ui}\rfloor}^{\alpha_{ui}^{\tau}}-k_{i2}^{\tau}{\lceil{e}_{Ui}\rfloor}^{\beta_{ui}^{\tau}}- {{\tilde g}_{ui}} and \dot{e}_{ri} = -k_{i3}^{\tau}{\lceil{e}_{ri}\rfloor}^{\alpha_{ri}^{\tau}}-k_{i4}^{\tau}{\lceil{e}_{ri}\rfloor}^{\beta_{ri}^{\tau}}- {{\tilde g}_{ri}} .
Lemma 1: The tracking error subsystem consisting of (4) and (7): [{e}_{qi}, {e}_{ri}] \mapsto [{e}_{pi}, {e}_{\psi i}] is input-to-state stable (ISS).
Proof: Choose the Lyapunov function as V_{i1}^e = \sum_{i = 1}^{N}(1/2)({e}_{pi}^T{e}_{pi}+ {e}_{\psi i}^2). The time derivative of V_{i1}^e is \dot{V}_{i1}^e\leq \sum_{i = 1}^{N} (-\lambda_{\max}(K_{i1})||H_{i1}||^2/ \sqrt{||H_{i1}||^2+\delta_{i \max}}+||h_{i1}||||H_{i1}||), where H_{i1} = [{e}_{pi}^T, {e}_{\psi i}] , K_{i1} = \text{diag}\{K_{qi}, K_{ri}\}, \delta_{i \max} = \max\{\delta_{i q}, \delta_{i r}\} , h_{i1} = [{e}_{qi}, {e}_{r i}] , and \lambda_{\min}(K_{i1}) is the minimum eigenvalue of a square matrix K_{i1} . Note that as ||H_{i1}||^2/ \sqrt{||H_{i1}||^2+\delta_{i \max}}\geq ||h_{i1}||/(a_{i1}\lambda_{\min}(K_{i1})), \dot{V}_{i1}^e\;\leq\; -\;\sum_{i = 1}^N\{\lambda_{\min}(K_{i1})(1- a_{i1})\}.
Lemma 2: Under Assumption 1, with the FTESO (10), the states \hat{U}_i , \hat{g}_{ui} , \hat{r}_{i} , and \hat{g}_{ri} can achieve the estimations in a fixed time.
Proof: Take the following part from error dynamics (11):
\begin{equation} \left\{ \begin{aligned} {\dot{\tilde{U}}}_i = &-k_{i1}^u {\lceil{\tilde U}_i\rfloor}^{\alpha_{u i}} + {{\tilde g}_{ui}} \\ {\dot{\tilde{g}}}_{ui} = &-k_{i3}^u {\lceil{\tilde U}_i\rfloor}^{2\alpha_{u i}-1}. \end{aligned} \right. \end{equation} | (14) |
Define {F}_{i1} = [\tilde{U}_{i},\tilde{{g}}_{ui}]^T and {\zeta}_{i1} = [\tilde{U}_{i}, ((\tilde{{g}}_{ui})^{1/\alpha_{ui}})]^T . Consider a Lyapunov function V_{i1}^u({\zeta}_{i1}) = {\zeta}_{i1} ^T P_{i1}{\zeta}_{i1} , where P_{i1} is a symmetric positive definite matrix satisfying P_{i1}A_{i1}+A_{i1}^TP_{i1} = -I and A_{i1} is a Hurwitz matrix with A_{i1} = [-k_{i1}^u,1;-k_{i3}^u,0] . Setting \alpha_{ui} = 1 , the equation (14) becomes \dot{F}_{i1} = A_{i1}{F}_{i1} . Taking the time derivatives of V_{i1}^u({F}_{i1}) gives \dot{V}_{i1}^u({F}_{i1}) = F_{i1}^T(P_{i1}A_{i1}+A_{i1}^TP_{i1})F_{i1}\; =\; -F_{i1}^TF_{i1}\;\leq\; 0. Define \sigma_{ui} as a small constant, and (14) is locally asymptotically stable by choosing \alpha_{ui} in the interval (1-\sigma_{ui},1) . Then, from Lemma 2 in [5], {V}_{i1}^u({\zeta}_{i1}) is homogeneous of degree \varrho_{i1} with (0,\alpha_{ui}) and its derivative \dot{V}_{i1}^u({\zeta}_{i1}) is homogeneous of degree \varrho_{i1}+\alpha_{ui}-1 . One has V_{i1}^u({F}_{i1})\leq \lambda_{\max}(P_{i})||F_{i1}||^2 and \dot{V}_{i1}^u({F}_{i1})\leq -||F_{i1}||^2 . It renders that \dot{V}_{i1}^u({F}_{i1})\leq -{V}_{i1}^u({F}_{i1})/\lambda_{\max}(P_{i1}) . The second part of (11) is expressed as {\dot{\tilde{U}}}_i = -k_{i2}^u {\lceil{\tilde U}_i\rfloor}^{\beta_{u i}} + {{\tilde g}_{ui}} and {\dot{\tilde{g}}}_{ui} = -k_{i4}^u {\lceil{\tilde U}_i\rfloor}^{2\beta_{u i}-1}. Define {\zeta}_{i2} = [\tilde{U}_{i}, ((\tilde{{g}}_{ui})^{1/\beta_{ui}})]^T . Consider a Lyapunov function V_{i2}({\zeta}_{i2}) = {\zeta}_{i2} ^T P_{i2}{\zeta}_{i2} . Similar to the analysis in (14), one has \dot{V}_{i2}({\zeta}_{i2})\leq -{V}_{i2}({\zeta}_{i2})^{(\varrho_{i2}+\beta_{ui}-1)/\varrho_{i2}}/ \lambda_{\max}(P_{i2}) . For
\begin{align} \left\{ {\begin{aligned} &{\dot{\tilde{U}}}_i = -k_{i1}^u {\lceil{\tilde U}_i\rfloor}^{\alpha_{ui}}-k_{i2}^u{\lceil{\tilde U}_i\rfloor}^{\beta_{ui}} + {{\tilde g}_{ui}} \\ &{\dot{\tilde{g}}}_{ui} = -k_{i3}^u {\lceil{\tilde U}_i\rfloor}^{2\alpha_{ui}-1}-k_{i4}^u{\lceil{\tilde U}_i\rfloor}^{2\beta_{ui}-1} \end{aligned} } \right. \end{align} | (15) |
there exists \iota_{ui} such that {V}_{i2}^u({\zeta}_{i2}(t_0))>\iota_{ui} with \iota_{ui} being 0<\iota_{ui}< \lambda_{\min}(P_{i2}) . It renders that {V}_{i2}^u reaches \iota_{ui} within a fixed-time T_{i1}^u = \varrho_{i2}\lambda_{\max}(P_{i2})/(\beta_{ui}-1)\iota_{ui}^{(\beta_{ui}-1)/\varrho_{i2}} . Besides, the inequalities ||{\zeta}_{i2}||^2\leq {V}_{i2}^u({\zeta}_{i2})/\lambda_{\min}(P_{i2})\leq \iota_{ui}/\lambda_{\min}(P_{i2})\leq 1 and {V}_{i1}^u({\zeta}_{i1})\leq \lambda_{\max}(P_{i1})||{\zeta}_{i1}||^2 hold. If {V}_{i1}^u({\zeta}_{i1})<\iota_{ui} and ||{\zeta}_{i1}|| = 1 , ||{\zeta}_{i1}||^2\leq ||{\zeta}_{i2}||^2\leq 1 and {V}_{i1}({\zeta}_{i1})\leq \lambda_{\max}(P_{i1}) . The settling time of the system (14) satisfies T_{i2}^u\leq \varrho_{i1} \lambda_{\max}^{(\varrho_{i1}+\alpha_{ui}-1)/\varrho_{i1}}(P_{i1})/(1-\alpha_{ui}) . As a result, the states of the system (15) are convergent to zero within a fixed time T_{i1}^u+T_{i2}^u . If t\geq T_{i1}^u+T_{i2}^u , the system (11) becomes {\dot{\tilde{g}}}_{ui} = k_{i0}^u\text{sgn}({{\tilde U}_i})-{{\dot g}_{ui}} . Consider a Lyapunov function {V}^u_{i3} =( 1/2)({\tilde{g}}_{ui})^2 . The derivative of {V}_{i3}^u can be obtained as \dot{V}_{i3}^u\leq -(k_{i0}^u-{g}^*_{ui})\sqrt{(2{V}_{i3}^u)} . In summary, {\tilde{g}}_{ui} is convergent to zero within the time T_{i3}^u\leq \sqrt{2{V}_{i3}^u(T_{i1}^u+T_{i2}^u)}/(k_{i0}^u-{g}^*_{ui}) . Thus, the states {\tilde{U}}_{i} and {\tilde{g}}_{ui} are convergent to zero within a fixed time T_{ui}\leq T_{i1}^u+T_{i2}^u+T_{i3}^u regardless of the initial conditions. Similarly, the states {\tilde{r}}_{i} and {\tilde{g}}_{ri} achieve the estimations within the fixed time T_{ri}\leq T_{i1}^r+T_{i2}^r+T_{i3}^r with T_{i1}^r = \varrho_{i4}\lambda_{\max}(P_{i4})/(\beta_{ri}-1)\iota_{ri}^{(\beta_{ri}-1)/\varrho_{i4}} , T_{i2}^r\leq \varrho_{i3} \lambda_{\max}^{(\varrho_{i3}+\alpha_{ri}-1)/\varrho_{i3}}(P_{i3})/(1-\alpha_{ri}), and T_{i3}^r\leq \sqrt{2{V}_{i3}^r(T_{i1}^r+T_{i2}^r)}/(k_{i0}^r-{g}^*_{ri}) regardless of the initial conditions. The definitions of P_{i2} , P_{i3} , P_{i4} , A_{i2} , A_{i3} , A_{i4} , \varrho_{i2} , \varrho_{i3} , \varrho_{i4} , and \iota_{ri} analogize the definitions of P_{i1} , A_{i1} , \varrho_{i1} , and \iota_{ui} .
Theorem 1: Under Assumption 1, by using the optimal total speed guidance law (5), the yaw guidance law (6), the nonlinear tracking differentiator (8), the fixed-time ESOs (10), and the anti-disturbance kinetic fixed-time control laws (12), the RSVs described by (2) are able to cover the centroid of the Voronoi cell (3) at any initial sates with the safe set {{\cal{C}}_{zi}} . Besides, all the errors in the closed-loop system are uniformly ultimately bounded.
Proof: According to Lemma 1, the tracking errors {e}_{qi} , {e}_{pi} , and {e}_{\psi i} are uniformly ultimately bounded (UUB). By [5], it can obtain that the errors e_{Ui}^c , e_{ri}^c , e_{Ui}^d , and e_{ri}^d can converge to a small neighborhood of the origin. Construct the Lyapunov function as V_{i4} =( 1/2)({e}_{Ui}^2+ {e}_{ri}^2). Taking the time derivative of the V_{i4} along (12), it can be obtained that \dot{V}_{i4}\;\leq \;-2^{(1+\alpha_{ui}^{\tau})/2}k_{i1}^{\tau}(1/2 |{e}_{Ui}|^2)^{1+\alpha_{ui}^{\tau}}\;-\; 2^{(1+\beta_{ui}^{\tau})/2}k_{i2}^{\tau} \times (1/2 |{e}_{Ui}|^2)^{1\;+\;\beta_{ui}^{\tau}}\;- 2^{(1+\alpha_{ri}^{\tau})/2}k_{i3}^{\tau}((1/2) |{e}_{ri}|^2)^{1\;+\;\alpha_{ri}^{\tau}}\;-\;2^{(1+\beta_{ri}^{\tau})/2}\;k_{i4}^{\tau}\;((1/ 2 )|{e}_{ri}|^2)^{1+\beta_{ri}^{\tau}}+ g_{ui}^*|{e}_{Ui}|+g_{ri}^*|{e}_{ri}|. Then, the following inequality holds \dot{V}_{i4}\leq -k_{i1}(V_{i4})^{\alpha_i}-2^{1-\beta_i}k_{i2}(V_{i4})^{\beta_i}+g_{ui}^*|{e}_{Ui}|+g_{ri}^*|{e}_{ri}| , where k_{i1}>0 , k_{i2}>0 , 1>\alpha_i>0 , and \beta_i>1 . It follows from Lemma 2 that the errors |{{\tilde g}_{ui}}| and |{{\tilde g}_{ri}}| converge to zero within a fixed-time and \dot{V}_{i4}\leq -k_{i1}(V_{i4})^{\alpha_i}-2^{1-\beta_i}k_{i3}(V_{i4})^{\beta_i} , where k_{i3} = 2^{1-\beta_i}k_{i2} . Thus, all the tracking errors in the entire closed-loop system are UUB.
Simulation: Consider a multi-RSV system including six RSVs. The environment has the xy dimensions of 100\;{\rm{m}}\times100\;{\rm{m}} and \phi(q) = 1 . The initial states of six charging station and six RSVs are set to p_{bi} = [2,85-15(i-1)]^T \; (i = 1,\ldots,6), p_{1} = [28,15]^T , p_{2} = [65, 43]^T, p_{3} = [37,20]^T , p_{4} = [12,15]^T , p_{5} = [55,50]^T , p_{6} = [51,12]^T , {\psi}_{b1} = {\psi}_{b6} = 0 , {\psi}_{b2} = \pi/2 , {\psi}_{b3} = -4\pi/5 , {\psi}_{b4} = \pi/4 , {\psi}_{b5} = \pi , q_{i} = [0,0]^T, {r}_{i} = 0 , E_1 = 3600\;{\rm{mV}} , E_2 = 3900\;{\rm{mV}} , E_3 = 4100\;{\rm{mV}} , E_4 = 3800\;{\rm{mV}} , E_5 = 4000\;{\rm{mV}} , and E_6 = 3700\;{\rm{mV}} . The parameters in this case are set to K_{ch} = 10 , K_{en} , k = 0.02 , d_{ch} , c = 0.5 , E_{\max} = 4200\;{\rm{mV}} , and E_{\min} = 3000\;{\rm{mV}} , k_{i1}^U = k_{i2}^U = k_{i1}^r = k_{i2}^U = 10 , k_{i3}^U = k_{i3}^r = 2.5 , k_{i1}^u = k_{i2}^u = k_{i1}^r = k_{i2}^r = 20 , k_{i3}^u = k_{i4}^u = k_{i3}^r = k_{i4}^r = 100 , k_{i0}^u = k_{i0}^r = 0.05 , \alpha_{ui} = \alpha_{ri} = 0.9 , and \beta_{ui} = \beta_{ri} = 1.1 , K_{qi} = \text{diag}\{1,1\} , k_{ri} = 0.5 , k_{i1}^\tau = k_{i2}^\tau = k_{i3}^\tau = k_{i4}^\tau = 1 , \alpha_{ui}^\tau = \alpha_{ri}^\tau = 0.9 , \beta_{ui}^\tau = \beta_{ri}^\tau = 1.1 , and \Delta_{qi} = \Delta_{ri} = 1 . The long duration coverage behavior and the energy change curve of the RSV battery are depicted in Fig. 1. It shows that the optimal coverage of the given coverage area is achieved for the first time at 125\;{\rm{s }}. At 220\;{\rm{s}} , the red RSV returns to the charging station for charging, and other RSVs continue the long duration coverage task. Next, the red RSV leaves the charging station to perform the coverage task. Then, at 630 s, the RSVs complete the optimal coverage for the second time.
Conclusion: In this letter, an anti-disturbance energy-aware control method for RSVs is proposed based on a CVT, two BE-CBFs, and FTESO to achieve long duration coverage task. By the control method, RSVs are capable of achieving the long duration coverage within the given coverage area.
Acknowledgments: This work was supported in part by the National Natural Science Foundation of China (51939001, 52301408), the National Science and Technology Major Project (2022ZD0119902), the Key Basic Research of Dalian (2023JJ11CG008), the Dalian Science and Technology Innovation Fund (2022JJ12GX034), and the Dalian Outstanding Young Scientific and Technological Talents Project (2022RY07).
[1] |
H. Li, M. Zhou, K. Hao, and E. Hou, “A survey of multi-robot regular and adversarial patrolling,” IEEE/CAA J. Autom. Sinica, vol. 6, no. 4, pp. 894–903, 2019. doi: 10.1109/JAS.2019.1911537
|
[2] |
C. Jie, S. Jian, and G. Wang, “From unmanned systems to autonomous intelligent systems,” Engineering, vol. 12, no. 5, pp. 16–19, 2022.
|
[3] |
T. Li, R. Zhao, C. Chen, L. Fang, and C. Liu, “Finite-time formation control of under-actuated ships using nonlinear sliding mode control,” IEEE Trans. Cybern., vol. 48, no. 11, pp. 3243–3253, 2018. doi: 10.1109/TCYB.2018.2794968
|
[4] |
L. Ma, Y.-L. Wang, and Q.-L. Han, “Cooperative target tracking of multiple autonomous surface vehicles under switching interaction topologies,” IEEE/CAA J. Autom. Sinica, vol. 10, no. 3, pp. 673–684, 2023. doi: 10.1109/JAS.2022.105509
|
[5] |
N. Gu, D. Wang, Z. Peng, and J. Wang, “Safety-critical containment maneuvering of underactuated autonomous surface vehicles based on neurodynamic optimization with control barrier functions,” IEEE Trans. Neural Netw. Learn. Syst., vol. 34, no. 6, pp. 2882–2895, 2023. doi: 10.1109/TNNLS.2021.3110014
|
[6] |
A. Kwok and S. Martinez, “Unicycle coverage control via hybrid modeling,” IEEE Trans. Automat. Contr., vol. 55, no. 2, pp. 528–532, 2010. doi: 10.1109/TAC.2009.2037473
|
[7] |
J. Li, Y. Xiong, J. She, and M. Wu, “A path planning method for sweep coverage with multiple UAVs,” IEEE Internet. Things J., vol. 7, no. 9, pp. 8967–8978, 2020. doi: 10.1109/JIOT.2020.2999083
|
[8] |
L. Zuo, W. Yan, R. Cui, and J. Gao, “A coverage algorithm for multiple autonomous surface vehicles in flowing environments,” Int. J. Control, vol. 14, no. 2, pp. 528–532, 2016.
|
[9] |
C. Song, L. Liu, G. Feng, Y. Fan, and S. Xu, “Coverage control for heterogeneous mobile sensor networks with bounded position measurement errors,” Automatica, vol. 120, p. 109118, 2016.
|
[10] |
J. Hou, X. Zeng, G. Wang, J. Sun, and J. Chen, “Distributed momentum-based frank-wolfe algorithm for stochastic optimization,” IEEE/CAA J. Autom. Sinica, vol. 10, no. 3, pp. 685–699, 2023. doi: 10.1109/JAS.2022.105923
|
[11] |
Z. Pang, W. Luo, G.-P. Liu, and Q.-L. Han, “Observer-based incremental predictive control of networked multi-agent systems with random delays and packet dropouts,” IEEE Trans. Circuits-II, vol. 68, no. 1, pp. 426–430, 2021.
|
[12] |
C. Zheng, Z. Pang, J. Wang, J. Sun, G.-P. Liu, and Q.-L. Han, “Null-space-based time-varying formation control of uncertain nonlinear second-order multi-agent systems with collision avoidance,” IEEE Trans. Ind. Electron., vol. 70, no. 10, pp. 10476–10485, 2023. doi: 10.1109/TIE.2022.3217585,2022
|
[1] | Zhaoyang He, Naiqi Wu, Rong Su, Zhiwu Li. Cyber-Attacks With Resource Constraints on Discrete Event Systems Under Supervisory Control[J]. IEEE/CAA Journal of Automatica Sinica, 2025, 12(3): 585-595. doi: 10.1109/JAS.2024.124596 |
[2] | Dan Zhang, Jiabin Hu, Jun Cheng, Zheng-Guang Wu, Huaicheng Yan. A Novel Disturbance Observer Based Fixed-Time Sliding Mode Control for Robotic Manipulators With Global Fast Convergence[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(3): 661-672. doi: 10.1109/JAS.2023.123948 |
[3] | Qinchen Yang, Fukai Zhang, Cong Wang. Deterministic Learning-Based Neural PID Control for Nonlinear Robotic Systems[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(5): 1227-1238. doi: 10.1109/JAS.2024.124224 |
[4] | Xuerao Wang, Qingling Wang, Yanxu Su, Yuncheng Ouyang, Changyin Sun. Adaptive Sensor-Fault Tolerant Control of Unmanned Underwater Vehicles With Input Saturation[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(4): 907-918. doi: 10.1109/JAS.2023.123837 |
[5] | Haijing Wang, Jinzhu Peng, Fangfang Zhang, Yaonan Wang. High-Order Control Barrier Function-Based Safety Control of Constrained Robotic Systems: An Augmented Dynamics Approach[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(12): 2487-2496. doi: 10.1109/JAS.2024.124524 |
[6] | Qian Ma, Peng Jin, Frank L. Lewis. Guaranteed Cost Attitude Tracking Control for Uncertain Quadrotor Unmanned Aerial Vehicle Under Safety Constraints[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(6): 1447-1457. doi: 10.1109/JAS.2024.124317 |
[7] | Li-Ying Hao, Gege Dong, Tieshan Li, Zhouhua Peng. Path-Following Control With Obstacle Avoidance of Autonomous Surface Vehicles Subject to Actuator Faults[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(4): 956-964. doi: 10.1109/JAS.2023.123675 |
[8] | Tangyike Zhang, Junxiang Zhan, Jiamin Shi, Jingmin Xin, Nanning Zheng. Human-Like Decision-Making of Autonomous Vehicles in Dynamic Traffic Scenarios[J]. IEEE/CAA Journal of Automatica Sinica, 2023, 10(10): 1905-1917. doi: 10.1109/JAS.2023.123696 |
[9] | Xiaohua Ge, Qing-Long Han, Qing Wu, Xian-Ming Zhang. Resilient and Safe Platooning Control of Connected Automated Vehicles Against Intermittent Denial-of-Service Attacks[J]. IEEE/CAA Journal of Automatica Sinica, 2023, 10(5): 1234-1251. doi: 10.1109/JAS.2022.105845 |
[10] | Zhaolei Wang, Qing Wang, Chaoyang Dong. Asynchronous H∞ Control for Unmanned Aerial Vehicles: Switched Polytopic System Approach[J]. IEEE/CAA Journal of Automatica Sinica, 2015, 2(2): 207-216. |
[11] | Chaoxu Mu, Qun Zong, Bailing Tian, Wei Xu. Continuous Sliding Mode Controller with Disturbance Observer for Hypersonic Vehicles[J]. IEEE/CAA Journal of Automatica Sinica, 2015, 2(1): 45-55. |
[12] | Yumei Li, Holger Voos, Mohamed Darouach, Changchun Hua. An Algebraic Detection Approach for Control Systems under Multiple Stochastic Cyber-attacks[J]. IEEE/CAA Journal of Automatica Sinica, 2015, 2(3): 258-266. |
[13] | Delong Hou, Qing Wang, Chaoyang Dong. Output Feedback Dynamic Surface Controller Design for Airbreathing Hypersonic Flight Vehicle[J]. IEEE/CAA Journal of Automatica Sinica, 2015, 2(2): 186-197. |
[14] | Lu Wang, Jianbo Su. Trajectory Tracking of Vertical Take-off and Landing Unmanned Aerial Vehicles Based on Disturbance Rejection Control[J]. IEEE/CAA Journal of Automatica Sinica, 2015, 2(1): 65-73. |
[15] | Derong Liu, Changyin Sun, Bin Xian. Guest Editorial for Special Issue on Autonomous Control of Unmanned Aerial Vehicles[J]. IEEE/CAA Journal of Automatica Sinica, 2015, 2(1): 1-1. |
[16] | Xiaoli Li, Kang Wang, Dexin Liu. An Improved Result of Multiple Model Iterative Learning Control[J]. IEEE/CAA Journal of Automatica Sinica, 2014, 1(3): 315-322. |
[17] | Jing Yan, Xian Yang, Cailian Chen, Xiaoyuan Luo, Xinping Guan. Bilateral Teleoperation of Multiple Agents with Formation Control[J]. IEEE/CAA Journal of Automatica Sinica, 2014, 1(2): 141-148. |
[18] | Ruimin Zhang, Lu Dong, Changyin Sun. Adaptive Nonsingular Terminal Sliding Mode Control Design for Near Space Hypersonic Vehicles[J]. IEEE/CAA Journal of Automatica Sinica, 2014, 1(2): 155-161. |
[19] | Kecai Cao, Bin Jiang, Dong Yue. Distributed Consensus of Multiple Nonholonomic Mobile Robots[J]. IEEE/CAA Journal of Automatica Sinica, 2014, 1(2): 162-170. |
[20] | Lijiao Wang, Bin Meng. Distributed Force/Position Consensus Tracking of Networked Robotic Manipulators[J]. IEEE/CAA Journal of Automatica Sinica, 2014, 1(2): 180-186. |