Professional Documents
Culture Documents
Abstract— This paper addresses the problem of high-speed velocity cannot be ignored due to the high speed of the vehicle.
waypoint tracking and real-time obstacle avoidance for Considering the above practical engineering problems, we design
large-scale underactuated autonomous underwater vehicles a robust nonlinear model predictive control (RNMPC) scheme
(AUVs) in the vertical plane. Specifically, a robust nonlinear from the kinematic level of the AUV to achieve high-speed way-
model predictive control (RNMPC) scheme is proposed, consid- point tracking and obstacle avoidance. And then we developed a
ering different types of constraints including the scale of the dynamic sensing and collision avoidance scheme to formulate
AUV, the finite field of view of the sensor, the input saturation, nonconvex collision avoidance into convex constraints in the
the physical limits on system state, and the influence of the RNMPC optimization problem. We demonstrate the effectiveness
vertical underactuated velocity. To navigate in the completely of the proposed control method in a high-fidelity simulation
unknown environment with nonconvex obstacles, a dynamic environment with large size terrain and shipwrecks as obsta-
sensing and collision avoidance scheme is proposed so that cles. In this environment, the AUV can dynamically read the
the collision avoidance can be properly formulated into convex distance data from obstacles to the sensor through the virtual
constraints in the RNMPC optimization problem. Recursive fea- multi-beam sonar equipped in the head of the vehicle, which
sibility and closed-loop stability are proved rigorously. Through can simulate the actual engineering experiment with high-fidelity.
the high-fidelity simulations with graph and data visualization This work could be applied to other underactuated systems,
techniques, the proposed algorithm has higher waypoint tracking such as unmanned aerial vehicles, unmanned surface vehicles,
accuracy, safer obstacle avoidance ability, and better multiple etc. In future research, we will consider more complex practical
constraints handling capability than the existing dynamic virtual working cases, such as the presence of dynamic obstacles in an
AUV (DVA) technique. unknown environment and the disturbance of current.
Note to Practitioners—This article was motivated by the Index Terms— Dynamic sensing, finite field of view, high-
problem of high-speed waypoint tracking and obstacle avoidance fidelity simulations, large-scale high-speed underactuated AUV,
for large-scale underactuated autonomous underwater vehicles obstacle avoidance, robust nonlinear MPC.
(AUVs) in an unknown environment. In practical engineering, the
obstacles’ useful information (e.g., shapes and positions) cannot I. I NTRODUCTION
be directly reflected by the raw data (distance or acoustic image)
of multi-beam sonars. And the multi-beam sonars are also limited
by detection range and angle. Another practical problem is that
U NDERACTUATED autonomous underwater vehicles
(AUVs) are widely applied in underwater missions due
to the mature technology and the advantages in energy and
such type of AUV suffers from different constraints including
the scale of the vehicle, the physical limits on system poses cost saving [1], [2], [3], [4]. At present, AUVs are develop-
and velocities. In addition, the influence of the underactuated ing towards large-scale and high speed to meet higher task
requirements (such as large carrying capacity, long range,
Manuscript received 27 April 2023; revised 22 December 2023; etc.). However, the large scale, high speed, underactuation,
accepted 23 February 2024. This article was recommended for publication
by Associate Editor L. Cheng and Editor P. Rocco upon evaluation of the
multiple constraints, and other complex characteristics of
reviewers’ comments. This work was supported in part by the National the vehicle bring huge challenges to underwater autonomous
Natural Science Foundation of China under Grant 52101365, in part by control design. In particular, a real-time high-speed obstacle
the LingChuang Research Project of China National Nuclear Corporation,
in part by the Young Talent Project of China National Nuclear Corporation,
avoidance system is the key enabling technology towards the
in part by the Young Elite Scientists Sponsorship Program by CAST under full autonomy of the AUV.
Grant 2021QNRC001, in part by the Shanghai Sailing Program under At present, real-time obstacle avoidance techniques for
Grant 21YF1419800, in part by the Oceanic Interdisciplinary Program of
Shanghai Jiao Tong University under Grant SL2021PT107, and in part by the the AUV mainly include rapidly-exploring random trees [5],
Shanghai Jiao Tong University 2030 Initiative under Grant WH510363001-6. artificial potential fields [6], fuzzy algorithms [7], neural
(Corresponding author: Lei Qiao.) networks [8], dynamic virtual AUV (DVA) principle [9],
Lin Yu and Lei Qiao are with the State Key Laboratory of Ocean
Engineering and the School of Naval Architecture, Ocean and Civil Engi- reinforcement learning and deep reinforcement learning [10],
neering, Shanghai Jiao Tong University, Shanghai 200240, China (e-mail: model predictive control (MPC) [11], and other algorithms.
yulin258@sjtu.edu.cn; qiaolei@sjtu.edu.cn). Since MPC algorithm has abilities to deal with multiple
Chao Shen is with the Department of Systems and Computer Engi-
neering, Carleton University, Ottawa, ON K1S 5B6, Canada (e-mail: constraints and reduce the impact of the large inertia [12],
shenchao@sce.carleton.ca). [13], it is much suitable for the obstacle avoidance problem
This article has supplementary material provided by the authors and of AUVs. In [11], a model predictive control based on particle
color versions of one or more figures available at
https://doi.org/10.1109/TASE.2024.3373607. swarm optimization and cubic spline interpolation is pro-
Digital Object Identifier 10.1109/TASE.2024.3373607 posed to realize the path tracking and obstacle avoidance, the
1545-5955 © 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
simulation results show the good performance of the proposed In addition, the large scale, high speed, and large inertia of the
strategy. Considering the effective detection distance of the AUV are also difficult problems to achieve obstacle avoidance
environment sensing sensor, Heshmati-Alamdari et al. [14], with real-time performance and security.
[15] propose nonlinear MPC and robust nonlinear MPC Motivated by the aforementioned considerations, this paper
(RNMPC) schemes for autonomous navigation of the under- designs an RNMPC controller for the large-scale underac-
actuated AUV operating with multiple constraints including tuated AUV to deal with the problem of waypoint tracking
static obstacles, workspace boundaries, and state and input and high-speed obstacle avoidance. Through the high-fidelity
constraints. To further reduce the conservatism of the approach simulation, the effectiveness and advantage of the proposed
in [15], a tube-based MPC is instead proposed in [16], where method are verified. The main contributions are as follows.
a tube is designed around the nominal trajectory of the system, 1) An RNMPC scheme is designed for a large-scale AUV
and is a priori guaranteed that the system will evolve in to realize high-speed waypoint tracking and real-time obstacle
real-time in a tube that is dependent on the disturbance. Con- avoidance in the vertical plane, respecting the state and input
sidering that the detection distance of the sensor is practically constraints and considering the influence of underactuated
short, a receding horizon optimization scheme combining path velocity. The recursive feasibility and closed-loop stability are
planning algorithm and MPC is proposed in [17] and [18]. rigorously proved.
The simulation work based on the real dynamic model of 2) Aiming at the problem that the detection distance and
the Falcon AUV demonstrates the effectiveness of the pro- range of real sensors are limited and the whole picture
posed control algorithm. To further optimize the performance of large-scale obstacles cannot be completely and directly
evaluation, the authors in [19] present a conditional-scenario detected by sensors, a dynamic environment sensing and
based MPC, allowing certain constraints to be violated with a approximate convex collision avoidance scheme is developed
guaranteed probability according to the actual scenario. Then to construct the obstacle avoidance constraints in the RNMPC
the original non-convex optimization problem for obstacle optimization problem.
constraint is transformed into multiple convex subproblems by 3) Compared with the obstacle avoidance algorithm based
introducing a novel concept of candidate paths. The optimal on the DVA principle proposed in [9], the RNMPC scheme
candidate path decides the input of the system. Simulation offers a safer obstacle avoidance capability, higher waypoint
results show that the proposed algorithm can successfully tracking accuracy, and better constraint processing ability
avoid obstacles with the probability of satisfying the priority than the DVA obstacle avoidance algorithm for high-speed
constraint. navigation of the large-scale underactuated AUV.
To our best knowledge, no study related to high-speed The remainder of this paper is organized as follows.
obstacle avoidance of large-scale underactuated AUVs has In Section II, the problem formulation is presented. Section III
been publicly reported, but the excellent work related to introduces the proposed control method with a rigorous
high-speed obstacle avoidance of unmanned aerial vehicles analysis of recursive feasibility and closed-loop stability.
(UAVs) [20], [21], [22], [23] has been continuously published. In Section IV, the effectiveness and advantage of the pro-
UAVs can sense the environment through optical sensors such posed algorithm are verified through high-fidelity simulations.
as lidar [20] and cameras [21], with fast information update Conclusive remarks are provided in Section V.
rate and long detection distance. In addition, the high mobility Notation: Throughout the paper, Rn denotes an
of UAVs also makes it possible to achieve high-speed obstacle n-dimensional real space and R≥0 denotes a non-negative real
avoidance. However, unlike UAVs, large-scale underactuated space. Vectors are represented in lowercase bold italics, such
AUVs do not have the long sensing range, fast information as x, matrices are shown in capital bold italics, such as A, and
update rate, and high maneuverability. Therefore, it is a big scalars are represented in fine italics, such as m. For a given
√
difficulty to ensure the safety of large-scale underactuated vector x, ẋ denotes its first-order derivative, ∥x∥2 = x T x,
AUVs in the process of high-speed navigation. ∥x∥2A = x T Ax, and x T is its own transpose. The maximum
The above studies about AUVs only consider that the and minimum eigenvalues of matrix Q are denoted by λ Q
perceptual area of the system is a circular region with a and λ Q , respectively. For two given sets A, B ⊆ Rn , A ∼
radius of the detectable distance, and assume that the obstacle B := {x ∈ Rn : x + y ∈ A,L∀ y ∈ B} denotes the Pontryagin
information (such as complete shape and position, etc.) can difference of sets. And A B := {x + y : x ∈ A, y ∈ B}
be obtained directly when there are obstacles in the perceptual represents Minkowski sum set. Moreover, B(c, r ) :=
range. The algorithm to properly formulate the needed sensing
x ∈ R2 : ∥x − c∥2 ≤ r, c ∈ R2 , r > 0 stands for a circle
information from raw data is missing. Another practical issue with center c and radius r .
is that the sensor’s field of view is a fan area instead of
a perfect circular area as assumed in most literature. The
II. P ROBLEM F ORMULATION
underactuated AUV considered in [17] and [18] is with a
very low speed, such that the underactuated speed can be A. Mathematical Modeling
approximately zero. However, for a high-speed AUV, the As seen in Fig. 1, the pose vector of the AUV with respect
vertical underactuated velocity cannot be ignored due to the to (w. r. t.) the
T inertial frame O E − X E Z E is defined
T by
imbalance between the gravity and buoyancy, lift force, and x = x z θ , including the position vector x z and
other reasons. If the influence of underactuated velocity is orientation θ. The velocity w. r. t. the body-fixed frame
T
not considered, the obstacle avoidance tends to fail in reality. O B − X B Z B is denoted by v = u w q , including linear
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
T
where J (x, u) = u cos θ −u sin θ q
is regarded as the
T
actuated term and K (x, w) = w sin θ w cos θ 0 the under-
T
actuated one, and u = u q is the input of the system.
Discretize equation (3) by difference, we can get
x k+1 = x k + J (x k , uk )1t + K (x k , wk )1t, (4)
where 1t is a sampling period, x k ∈ X ⊂ R and uk ∈ 3
tracking task, Pl−1 is the previous waypoint with coordinates Define the vector x̂(k + i|k) as the predicted state of the
(xl−1 , zl−1 ) and Pl is the current one with the coordinates system at sample time k + i according to the actual sys-
(xl , zl ), where l (l = 1, 2, · · · , L) respects lth waypoint, tem state x k at time k applying a control sequence uk =
and L represents the total number of waypoints. The attitude {u(k|k), u(k + 1|k), · · · , u(k + i − 1|k)} using the actuated
of the line between the two points can be expressed in the model. Hence,
inertial frame as
x̂(k + i|k) = f x̂(k + i − 1|k), u(k + i − 1|k) (6)
zl − zl−1
θl = −arctan . (1) with x̂(k|k) = x k .
xl − xl−1
Property 1: The actuated model f (x, u) of the system is
When the AUV enters the red circle with radius Rd , it is locally Lipschitz in X , that is, for every u ∈ U, such that
considered as reaching the current waypoint, and then it can
continue to track the next one. ∥ f (x 1 , u) − f (x 2 , u)∥2 ≤ L f ∥x 1 − x 2 ∥2 (7)
The kinematic equations of the AUV in the vertical plane
with L f = 1 + u1t and u > 0 is the upper bound of u.
can be given as [24] Proof: Assuming that x 1 , x 2 ∈ X , then
ẋ = u cos θ + w sin θ
∥ f (x 1 , u) − f (x 2 , u)∥2
ż = −u sin θ + w cos θ (2)
x1 − x2
u cos θ1 1t − u cos θ2 1t
θ̇ = q.
≤ z 1 − z 2 + −u sin θ1 1t + u sin θ2 1t
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
And the obstacles on the seafloor should not be too high to Similarly, equation (16) can be proved by
exceed the AUV’s maximum obstacle avoidance capacity in
such a way that there exists a feasible solution of optimization ∥V (x 1 ) − V (x 2 )∥2 ≤ 2ελ P ∥x 1 − x 2 ∥2
problem under the constraints of pitch and its velocity. Based = L v ∥x 1 − x 2 ∥2 .
on the above considerations, the state constraint set X and Assumption 3 There exists a local controller uck = κ(x) ∈
velocity constraint set U can be given by U, and for ∀x k ∈ 8, the following stability condition is
X := x ∈ R3 : (12), |θ| ≤ θ ,
(13) satisfied
U := u ∈ R2 : 0 ≤ u ≤ u, |q| ≤ q ,
(14) Now, we will design the terminal constraint set to
satisfy Assumption 3. Let u∗ (k − 1) be the feasible optimal
where θ̄ is the upper bound of θ with 0 < θ < π/2, q > 0 control sequence obtained by the optimization problem in
is the upper bound of the pitch velocity q. It is worth noting equation (9a)-(9f) at time k − 1. Next, the feasible control
that 0 < θ < π/2 could ensure that the kinematics described sequence of the optimization problem at time k will be
by Eular angles are nonsingular. defined as
The design of the terminal constraint will directly affect the
(
u∗ (k + i | k − 1) i = 0, · · · , N − 2
stability of the system. To ensure the stability of the system, ũ(k + i | k) =
some assumptions should be made as follows. κ(x) i = N − 1.
Assumption 2 For an actuated system (5), there exists an (18)
admissible positively invariant set 8 := x ∈ R3 : x −
x l 2 ≤ ε, ε > 0 ⊂ X , such that the terminal constraint Let
Xteminal ⊂ 8. κ(x) = K (x − x l ), (19)
Lemma 2 The running cost function satisfies L(x l , 0) = 0,
and there exists a positive parameter c > 0, such that where
L(x, u) ≥ c∥x − x l ∥22 . L(x, u) is Lipschitz continuous in X ,
k1 0 0
i.e. for all x 1 , x 2 ∈ X and all u ∈ U, K=
0 0 k2
∥L(x 1 , u) − L(x 2 , u)∥2 ≤ L c ∥x 1 − x 2 ∥2 . (15) with k1 , k2 < 0, and P = diag{0.5, 0.5, 0.5}, then we
calculate the following
And the terminal cost function V (x) is also Lipschitz contin-
uous in 8, i.e. for all x 1 , x 2 ∈ 8, V ( f (x k , uck ))−V (x k ) + L(x k , uck )
1 2 2
∥V (x 1 ) − V (x 2 )∥2 ≤ L v ∥x 1 − x 2 ∥2 . (16) = k1 △t + q1 + r1 k1 2 + cos θ △ tk1 (xk − xl )2
2
Proof:
1 2 2
+ k2 △t + q3 + r2 k2 + △tk2 (θk − θl )2
2
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
at the same time. Next, an obstacle is taken as an example to Based on the above, the position constraints of the AUV
establish its collision model. Assuming that the sensor returns can be designed in detail as
b ≥ 2 valid data, the index value of the effective starting
dist B( p, R), B po , ro ≥ dmin ,
beam is a, that is, the valid data is {da , · · · , da+b−1 }. The (31)
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
where dmin > 0 is the safe distance, which can further ensure
the security of the AUV during high-speed movement. In this
case, the state constraint can be summarized as follows.
X = x ∈ R3 : dist B( p, R), B pom , rom
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Fig. 4. The visualized simulation results of the obstacle avoidance process. Figures (left to right, top to bottom) show the results at 49.8s (a), 82.4s (b),
107.9s (c), 154.7s (d), 167.9s (e) and 207s (f), respectively. The green fan on the head of the vehicle is a visualization of the multi-beam sonar. The green
line along the Z B -axis of the body-fixed frame is the visualization of the sonar altimeter. The green circle represents the circular neighborhood with a radius
of 5 meters around the waypoint.
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
R EFERENCES
[1] J. Yuh, “Design and control of autonomous underwater robots:
A survey,” Auton. Robots, vol. 8, no. 1, pp. 7–24, 2000.
[2] F. Zhang, G. Marani, R. N. Smith, and H. T. Choi, “Future trends in
Fig. 6. The height of the AUV from the obstacle directly below it. The green
marine robotics,” IEEE Robot. Autom. Mag., vol. 22, no. 1, pp. 14–122,
dashed line indicates the preset safe distance dmin .
Mar. 2015.
[3] Y. Shi, C. Shen, H. Fang, and H. Li, “Advanced control in marine
mechatronic systems: A survey,” IEEE/ASME Trans. Mechatronics,
vol. 22, no. 3, pp. 1121–1131, Jun. 2017.
[4] M. Cai, Y. Wang, S. Wang, R. Wang, Y. Ren, and M. Tan, “Grasp-
ing marine products with hybrid-driven underwater vehicle-manipulator
system,” IEEE Trans. Autom. Sci. Eng., vol. 17, no. 3, pp. 1443–1454,
Jul. 2020.
[5] E. Taheri, M. H. Ferdowsi, and M. Danesh, “Closed-loop randomized
kinodynamic path planning for an autonomous underwater vehicle,”
Appl. Ocean Res., vol. 83, pp. 48–64, Feb. 2019.
[6] Y. Noguchi and T. Maki, “Path planning method based on artificial
Fig. 7. The pitch angle of the AUV and its constraints. The dashed green potential field and reinforcement learning for intervention AUVs,” in
lines are the constraints of the pitch angle θ . Proc. IEEE Underwater Technol. (UT), Kaohsiung, Taiwan, May 2019,
pp. 1–6.
[7] M. Fang, S. Wang, M. Wu, and Y. Lin, “Applying the self-tuning fuzzy
control with the image detection technique on the obstacle-avoidance
for autonomous underwater vehicles,” Ocean Eng., vol. 93, pp. 11–24,
Sep. 2015.
[8] X. Cao and J. Peng, “A potential field bio-inspired neural network
control algorithm for AUV path planning,” in Proc. IEEE Int. Conf.
Inf. Autom. (ICIA), Wuyishan, China, Aug. 2018, pp. 1427–1432.
[9] J. Liu, M. Zhao, and L. Qiao, “Adaptive barrier Lyapunov function-
based obstacle avoidance control for an autonomous underwater vehicle
with multiple static and moving obstacles,” Ocean Eng., vol. 243,
pp. 110303–110318, Jan. 2022.
Fig. 8. The pitch velocity of the AUV with the constraints. The dashed green
[10] C. Zhang, P. Cheng, B. Du, B. Dong, and W. Zhang, “AUV path
lines are the constraints of q.
tracking with real-time obstacle avoidance via reinforcement learning
under adaptive constraints,” Ocean Eng., vol. 256, pp. 111453–111463,
Jan. 2022.
DVA in the complex and unknown environment. From Fig. 7, [11] X. Wang, X. Yao, and L. Zhang, “Path planning under constraints
and path following control of autonomous underwater vehicle with
it is observed that under both algorithms the pitch attitude dynamical uncertainties and wave disturbances,” J. Intell. Robot. Syst.,
constraints are not violated. However, it does not mean the vol. 99, nos. 3–4, pp. 891–908, Sep. 2020.
DVA can deal with the constraint. This can be verified in the [12] S. Kong, J. Sun, C. Qiu, Z. Wu, and J. Yu, “Extended state observer-
based controller with model predictive governor for 3-D trajectory
control input curves presented in Fig. 8, in which the pitch tracking of underactuated underwater vehicles,” IEEE Trans. Ind. Infor-
velocity produced by RNMPC can well satisfy its constraint mat., vol. 17, no. 9, pp. 6114–6124, Sep. 2021.
the whole time, whereas DVA does not. This is mainly because [13] J. Wang, Z. Wu, M. Tan, and J. Yu, “Model predictive control-
based depth control in gliding motion of a gliding robotic dolphin,”
the DVA does not have the capability to handle the constraint IEEE Trans. Syst., Man, Cybern., Syst., vol. 51, no. 9, pp. 5466–5477,
problem. Dec. 2021.
Video: A video demonstrating the simulation scenarios of [14] S. Heshmati-Alamdari, G. C. Karras, and K. J. Kyriakopoulos,
“A predictive control approach for cooperative transportation by multiple
this section can be found in the following links: https://sjtu- underwater vehicle manipulator systems,” IEEE Trans. Control Syst.
mirus.github.io/MIRUS.github.io/research/2024-02-29-xlauv1 Technol., vol. 30, no. 3, pp. 917–930, May 2022.
or https://www.bilibili.com/video/2024-02-29-xlauv1. [15] S. H. Alamdari, G. C. Karras, P. Marantos, and K. J. Kyriakopoulos,
“A robust predictive control approach for underwater robotic vehicles,”
IEEE Trans. Control Syst. Technol., vol. 28, no. 6, pp. 2352–2363,
V. C ONCLUSION Nov. 2020.
[16] S. Heshmati-Alamdari, A. Nikou, and D. V. Dimarogonas, “Robust
In this paper, we propose an RNMPC method for a trajectory tracking control for underactuated autonomous underwater
large-scale underactuated AUV to solve the problem of vehicles in uncertain environments,” IEEE Trans. Autom. Sci. Eng.,
vol. 18, no. 3, pp. 1288–1301, Jul. 2021.
high-speed waypoint tracking and real-time obstacle avoidance
[17] C. Shen, Y. Shi, and B. Buckham, “Model predictive control for an
in a vertical plane, considering different types of constraints AUV with dynamic path planning,” in Proc. 54th Annu. Conf. Soc.
including the state physical limitation, input saturation, and Instrum. Control Engineers Jpn. (SICE), Hangzhou, China, Jul. 2015,
the influence of underactuated velocity. Under a finite field of pp. 475–480.
[18] C. Shen, Y. Shi, and B. Buckham, “Integrated path planning and tracking
view of multi-beam sonar, a dynamic sensing and collision control of an AUV: A unified receding horizon optimization approach,”
avoidance scheme is developed to formulate the collision IEEE/ASME Trans. Mechatronics, vol. 22, no. 3, pp. 1163–1173,
avoidance into convex constraints in the RNMPC optimiza- Jun. 2017.
[19] X. Shang, J. Chen, S. Zhuang, and Y. Shi, “Scenario-based model
tion problem. From the high-fidelity simulation results, it is predictive control for path planning and obstacle avoidance,” in Proc. 4th
shown that the proposed control method can provide higher IEEE Int. Conf. Ind. Cyber-Phys. Syst. (ICPS), May 2021, pp. 446–451.
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
[20] F. Kong, W. Xu, Y. Cai, and F. Zhang, “Avoiding dynamic small Lei Qiao (Member, IEEE) received the B.S. degree
obstacles with onboard sensing and computation on aerial robots,” IEEE in automation and the M.Eng. degree in control
Robot. Autom. Lett., vol. 6, no. 4, pp. 7869–7876, Oct. 2021. engineering from the College of Automation, Harbin
[21] Z. Ma, C. Wang, Y. Niu, X. Wang, and L. Shen, “A saliency-based Engineering University, Harbin, China, in 2012 and
reinforcement learning approach for a UAV to avoid flying obstacles,” 2014, respectively, and the Ph.D. degree in con-
Robot. Auto. Syst., vol. 100, pp. 108–118, Feb. 2018. trol science and engineering from the Department
[22] D. Wang, T. Fan, T. Han, and J. Pan, “A two-stage reinforcement learning of Automation, Shanghai Jiao Tong University,
approach for multi-UAV collision avoidance under imperfect sensing,” Shanghai, China, in 2020.
IEEE Robot. Autom. Lett., vol. 5, no. 2, pp. 3098–3105, Apr. 2020. From September 2017 to September 2018, he was
[23] D. Huo, L. Dai, R. Xue, and Y. Xia, “Collision-free model predictive a Visiting Research Scholar with the Department
trajectory tracking control for UAVs in obstacle environment,” IEEE of Electrical and Computer Engineering, The Ohio
Trans. Aerosp. Electron. Syst., vol. 59, no. 3, pp. 5466–5477, Dec. 2022, State University, Columbus, OH, USA. He is currently an Assistant Professor
doi: 10.1109/TAES.2022.3221702. with the School of Naval Architecture, Ocean and Civil Engineering and the
[24] T. I. Fossen, Handbook of Marine Craft Hydrodynamics and Motion State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University.
Control. Hoboken, NJ, USA: Wiley, 2011. His research interests include navigation, control, coordination and games of
[25] O. Elhaki and K. Shojaei, “Neural network-based target tracking control marine robotics, and autonomous systems.
of underactuated autonomous underwater vehicles with a prescribed Dr. Qiao received the 2019 Premium Award for Best Paper in IET Control
performance,” Ocean Eng., vol. 167, pp. 239–256, Nov. 2018. Theory and Applications as the first author and the Young Scientist Award
[26] D. L. Marruedo, T. Alamo, and E. F. Camacho, “Input-to-state stable of 2022 International Conference on Automation, Control, and Robotics
MPC for constrained discrete-time nonlinear systems with bounded addi- Engineering. He was named in Stanford University List of Top 2% Scientists
tive uncertainties,” in Proc. 41st IEEE Conf. Decis. Control, Jun. 2002, Worldwide in 2023.
pp. 4619–4624.
[27] Z.-P. Jiang and Y. Wang, “Input-to-state stability for discrete-time
nonlinear systems,” Automatica, vol. 37, no. 6, pp. 857–869, 2001.
[28] M. Quigley, “ROS: An open-source robot operating system,” in Proc.
ICRA Workshop Open Source Softw., 2009.
[29] M. Prats, J. Pérez, J. J. Fernández, and P. J. Sanz, “An open source
tool for simulation and supervision of underwater intervention mis-
Chao Shen (Member, IEEE) received the B.E.
sions,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., Oct. 2012,
degree in automation engineering and the M.Sc.
pp. 2577–2582.
degree in control science and engineering from
[30] MS400U All-in-One Multibeam Echo Sounder. Accessed: Apr. 27, 2023. Northwestern Polytechnical University, Xi’an,
[Online]. Available: http://www.hydro-tech.cn/en/ms400u.html China, in 2009 and 2012, respectively, and the
[31] J. A. E. Andersson, J. Gillis, G. Horn, J. B. Rawlings, and M. Diehl, Ph.D. degree in mechanical engineering from
“CasADi: A software framework for nonlinear optimization and optimal the University of Victoria, Victoria, BC, Canada,
control,” Math. Program. Comput., vol. 11, no. 1, pp. 1–36, Mar. 2019. in 2018.
He is currently an Assistant Professor with the
Department of Systems and Computer Engineering,
Lin Yu received the B.E. degree in automation from Carleton University, Ottawa, ON, Canada. Before
Shandong University, Shandong, China, in 2018, joining Carleton University, he was a Post-Doctoral Researcher with
and the M.Sc. degree in control science and engi- the Real-Time Adaptive Control Engineering Laboratory, University of
neering from Tianjin University, Tianjin, China, Michigan, Ann Arbor, MI, USA. His research interests include control
in 2021. He is currently pursuing the Ph.D. degree theory, machine learning, and optimization, and their applications in robotics
in naval architecture and ocean engineering with systems, mechatronics systems, and industrial processes. He was a recipient
the State Key Laboratory of Ocean Engineering, of the 2018 IEEE SMCS Thesis Grant Initiative for the Ph.D. thesis on
Shanghai Jiao Tong University, Shanghai, China. model predictive control for underwater robotics and the Natural Science
His research interests include path planning, con- and Engineering Research Council of Canada Postdoctoral Fellowship in
trol theory, and their applications in unmanned 2020. He is also an Associate Editor of IEEE/ASME T RANSACTIONS ON
autonomous systems, such as autonomous underwa- M ECHATRONICS and IEEE C ANADIAN J OURNAL OF E LECTRICAL AND
ter vehicles, unmanned aerial vehicles, and autonomous ground vehicles. C OMPUTER E NGINEERING.
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 29,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.