Elsevier

Robotics and Autonomous Systems

Volume 87, January 2017, Pages 81-103
Robotics and Autonomous Systems

An autonomous reactive architecture for efficient AUV mission time management in realistic dynamic ocean environment

https://doi.org/10.1016/j.robot.2016.09.007Get rights and content

Abstract

Today AUVs operation still remains restricted to very particular tasks with low real autonomy due to battery restrictions. Efficient path planning and mission scheduling are principle requirement toward advance autonomy and facilitate the vehicle to handle long-range operations. A single vehicle cannot carry out all tasks in a large scale terrain; hence, it needs a certain degree of autonomy in performing robust decision making and awareness of the mission/environment to trade-off between tasks to be completed, managing the available time, and ensuring safe deployment at all stages of the mission. In this respect, this research introduces a modular control architecture including higher/lower level planners, in which the higher level module is responsible for increasing mission productivity by assigning prioritized tasks while guiding the vehicle toward its final destination in a terrain covered by several waypoints; and the lower level is responsible for vehicle’s safe deployment in a smaller scale encountering time-varying ocean current and different uncertain static/moving obstacles similar to actual ocean environment. Synchronization between higher and lower level modules is efficiently configured to manage the mission time and to guarantee on-time termination of the mission. The performance and accuracy of two higher and lower level modules are tested and validated using ant colony and firefly optimization algorithm, respectively. After all, the overall performance of the architecture is investigated in 10 different mission scenarios. The analysis of the captured results from different simulated missions confirms the efficiency and inherent robustness of the introduced architecture in efficient time management, safe deployment, and providing beneficial operation by proper prioritizing the tasks in accordance with mission time.

Introduction

Most of the current AUVs applications are supervised from the support vessel that provides higher-level decisions in critical situation and generally takes enormous cost during the mission [1]. A growing attention has been devoted in recent years on increasing the ranges of missions, vehicles endurance, extending vehicles applicability, promoting vehicles autonomy to handle longer missions without supervision, and reducing operation costs [2]. AUVs should operate in turbulent undersea environment with complex spatio-temporal variability, where the current variability can perturb AUVs safety conditions. Current instabilities can influence vehicle’s path and probably push it to an undesired direction. Robustness of AUV to this strong environmental variability is a key element to mission performance and carry out safety considerations. Restrictions of a priori knowledge about later conditions of the environment reduce AUVs autonomy and robustness. Additionally, AUVs capabilities in handling mission objectives are directly influenced by performance of task allocation system. Many of the today’s AUV missions are limited to executing a list of pre-programmed instructions and completing a predefined sequence of tasks. Autonomous adaptation of AUVs in performing different tasks in dynamic and continuously changing environment has not been completely fulfilled yet and it is still necessary for the operators to remain in the loop of considering and making decisions [3].

Generally, the AUVs control architecture consists of two different execution layers based on its control requirements i.e. deliberative and reactive layers [4]. The deliberative layer manages the concurrent execution of several tasks with different priorities. The reactive layer manages real-time reactions to perform quick response to critical events. These automatic functions are executed in the background and promote AUVs self-management characteristics. Mission timing and AUV’s time management are fundamental requirements toward mission success. Consequently, reaching higher performance in real-time applications is a challenging issue due to limited resource availability and the dynamic changes of the environment. Thus, the vehicle should trade-off within the problem constraints and mission productivity while managing the available time and risks. Evidently, decision autonomy and vehicles routing/task assignment capability are connected in various standpoints. On the other hand, instantly adapting to the continuously changing situations of the real world where most of the information are uncertain and unknown is strongly critical. Accordingly, in lower level also, the vehicle should be capable of copying dynamic time variant ocean current and avoiding collision to different static and dynamic obstacles to ensure safe deployment during the mission. To satisfy the addressed requirements and handling these challenges, development of more evolved embedded functions is required, which can promote the platform’s autonomy in both higher and lower levels while maintaining the trust on vehicles safety at all stages of the mission. Many efforts have been devoted in recent years for promoting AUVs’ autonomy in mission task assignment in recent years; however, there are still many challenges on achieving a satisfactory level of autonomy for AUV in this regard. Very few investigations are found to address issues with both low and high level autonomy development and the problem remains unsolved for underwater missions. According to these fundamental concepts, various methodologies have been discussed and worked out in the literature as follows.

Effective routing has a great impact on vehicles time management as well as mission performance by appropriate selection and arrangement of the tasks sequence. Liu and Bucknall (2015) proposed a three-layer structure to facilitate multiple unmanned surface vehicles to accomplish task management and formation route planning in a maritime environment [5]. Eichhorn (2015) implemented graph-based methods for the AUV “SLOCUM Glider” motion planning in a dynamic environment [6]. The author employed modified Dijkstra Algorithm where the applied modification and conducted time variant cost function simplifies the determination of a time-optimal route in the geometrical graph. An energy efficient fuzzy based route planning system is presented by Kladis et al., (2011) for UAVs motion planning in a graph-like terrain using a priori known wind information [7]. Other methods also studied on efficient task assignment for single/multiple vehicle task assign/routing problem such as graph matching algorithm [8], partitioning method [9], Tabu search algorithm [10], branch and cut algorithm [11], A* search algorithm [12], [13], [14], Dijkstra [6], [15], and evolutionary algorithms [8], [16], [17]. The traditional algorithms used for graph routing problem such as Dijkstra, or dynamic programming algorithm has major shortcomings such as high computational complexity for real-time applications. Another issue is that all outline literature in this scope mostly focus on routing problem for different unmanned vehicles in which task allocation is the principle direction of these papers and quality of deployment (motion) has not been addressed. The vehicle’s safe and confident deployment is a critical factor affecting vehicle’s level of autonomy and should be taken into consideration at all stages of the mission in a vast and uncertain environment. So, in the second category, existing AUVs’ trajectory/path planning approaches have been discussed, which are more concentrated on quality of vehicle’s motion encountering dynamicity of the terrain.

Path planning itself is a complicated multi-objective Non-deterministic Polynomial-time (NP) hard problem that has a great impact on vehicle’s overall autonomy. The path planning strategies proposed in most of the previous researches particularly deal with vehicle’s guidance toward the destination encountering dynamic changes of the terrain. Different methods like D*, A*, FM, FM* algorithms have been employed for AUV optimum path generation [18], [19], [20], [21]. Particularly, the main drawback of all above-mentioned methods is that their time complexity increases exponentially with increasing the problem space, which is inappropriate for real-time applications. Evolutionary algorithms are population based optimization methods applied successfully on path planning problem that are advantaged to be implemented on a parallel machine with multiple processors, which speeds up the computation process [22], [23]. The Particle Swarm Optimization (PSO) [22], [24], Evolution (DE) [24], [25], Genetic Algorithm (GA) [22], [26], [27], multi-objective genetic algorithm (NSGA II) [28], Differential], and Quantum-based PSO (QPSO) [29] are some popular types of optimization algorithms applied successfully on offline/online path planning approaches. Although various path planning techniques have been suggested for autonomous vehicles, AUV-oriented applications still face several difficulties. Path replanning in this context requires a significant computational effort for progressing data from update of entire terrain, which is problematic in large scale operations. Accurate prediction of the behavior of a dynamic large-scale terrain, far-off the vehicles sensor coverage is unreliable and impractical as only awareness of environment in vicinity of the vehicle is sufficient such that the vehicle can be able to perform reaction to prompt changes. Another problem is that the path planning strategies are not designed for handling the task assignment and time management in a graph-like terrain with respect to graph routing restrictions in cases that the vehicle is required to carry out a specific sequence of ordered tasks. In these kind of problems, a routing strategy is a proper approach in dealing with graph search constraints for the VRP problem and organizing the task. However, routing strategies are not accurate enough in copping dynamic variations of the terrain, but these approaches fracture the operation area to smaller beneficent sections and give overview of the pathway for vehicle’s maneuver that helps reducing the computational load. Accordingly, each approach is able to handle only a specific level of autonomy in terms of organizing the tasks and managing the mission or dealing with variations of the environment and producing a high quality maneuver for the vehicle. This research addresses the AUV’s requirements in both high and low level autonomy toward making AUV systems more intelligent and robust in managing its availabilities and adaption to a dynamic environment.

A novel autonomous architecture with online re-planning capability is developed to carry out the underwater missions in a large scale environment in the presence of severe environmental disturbances. The system incorporates two different execution layers, deliberative and reactive, to satisfy the AUV’s high and low level autonomy requirements in path planning and mission management. A “Task Organize Mission Plan” (TOM-P) module is designed for the deliberative layer to provide a higher level of autonomy by ordering the execution of several tasks so that the AUV gets directed toward the targeted location. The TOM-P module in the top level simultaneously rearranges the order of tasks taking passage of time into account. The reactive layer, on the other hand, is responsible for performing real-time reactions and quick response to critical events. A “Local On-line Path-Plan” (LOP-P) module was developed in this layer and includes automatic functions that executed in the background to promote the AUV’s self-management characteristics. The LOP-P module at the lower level deals with quality and safety of deployment along the ordered tasks (produced by higher layer TOM-P) where persistent variation of the sub-area in proximity of the vehicle is considered simultaneously. The operation of each module may take longer than expectation due to unexpected dynamic changes in the environment. In order to reclaim the missed time an efficacious synchronization scheme (named “Synchron”) is added to the architecture to keep pace of modules in different layers of the system.

To handle the complexity of NP-hard graph routing and task allocation problem, the TOM-P module utilizes the Ant Colony Optimization (ACO) algorithm to find an optimum order of tasks for the underwater mission. In the LOP-P module, Firefly Algorithm (FFA) is conducted to carry out path planning between each pair of the waypoints, which is efficient and fast enough in generating collision-free optimum trajectory in smaller scale. The argument for using the ACO in solving vehicle routing and task assignment problems is strong enough due to its discrete nature and strong capability in scaling well with complex problems [40], [41], [42]. On the other hand, the FFA is advantaged to use an automatic subdivision approach that makes it specifically suitable and flexible in dealing with continuous problems (such as path planning), highly nonlinear problems, and multi-objective problems [45], [46]. This fact increases convergence rate of the algorithm. Moreover, its control parameters can be tuned iteratively that increases convergence rate of the algorithm.

This research is a completion of the previous attempts [30], [31], [32] that construction of the mission-path planners [31], [32] is boosted to a consistent structure in scheme of a coherent architecture in which the performance of the proposed construction is mainly independent of the employed algorithms by modules. The main reason for remarkable performance of this system is the fashion of mixing and matching two disparate strategies from two different perspectives and constructing an accurate synchronization between them. A significant benefit of such modular construction is that the modules can employ different methods or their functionality can get upgraded without manipulating the system’s structure. This advantage specifically increases the reusability and versatility the control architecture and eases updating/upgrading AUV’s functionalities to be compatible with other applications, and in particular implement it for other autonomous systems such as unmanned aerial, ground or surface vehicles. The previous study [30], which provided a primary basic idea of such a modular framework, is expanded by enhancing the synchronization process and maneuverability of the vehicle by upgrading the path planner with the added capability of reactive re-planning, generalizing the applicability of the planners by modeling more realistic underwater situations incorporating kinodynamics of the AUV, variable water currents and uncertainty of the terrain.

The paper is prepared in following sections: the mechanism of the TOM-P and LOP-P modules are demonstrated in Section 2, respectively. An overview of the adopted methods by both of modules on corresponding problems is presented in Section 3. The architectures evaluation and analysis of the simulation results is provided in Section 4 and the conclusion of this research is provided by Section 5.

Section snippets

Proposed modular control architecture

Proposed architecture designed in separate modules running concurrently including TOM-P in top level with higher level of decision autonomy, and the LOP-P in lower level to autonomously carry out the collision avoidance, cope with current force and handle similar environmental challenges. The modules interact simultaneously by back feeding the situational awareness of the surrounding operating field; accordingly, the system decides to re-plan the path or mission or continue the current mission.

Methodology adopted by higher/lower level modules of the architecture

There is a significant distinction between theoretical understanding of meta-heuristics and peculiarity of different applications in contrast. Specifically, this gap is more highlighted when scale (size), complexity, and nature of the problem is taken to account. Application of different methods may result very diverse on a same problem due to specific nature each problem. To handle the complexity of NP-hard graph routing and task allocation problem, the TOM-P module utilizes the ACO algorithm

Discussion and analysis of simulation results

To validate the architecture’s performance, first higher and lower level modules are evaluated separately, then their synchronous collaboration is investigated and evaluated. For the purposes of this study, the optimization problem was performed on a desktop PC with an Intel i7 3.40 GHz quad-core processor in MATLAB® R2016a.

Conclusion and future works

Generally, the AUV’s control architecture consists of two different execution layers based on its control requirements i.e. deliberative and reactive layers. The deliberative layer manages the concurrent execution of several tasks with different priorities. The reactive layer carries out the real-time reactions to perform quick response to critical sudden events. These automatic functions are executed in the background and promote AUVs self-management characteristics. To provide a higher level

Somaiyeh Mahmoud Zadeh completed her Master degree in Computer Science at National University of Malaysia. She is currently Ph.D. student at School of Computer Science, Engineering and Mathematics, Flinders University of South Australia. Her area of research includes computational intelligence, autonomy and decision making, situational awareness, and motion planning of autonomous underwater vehicles.

References (47)

  • FurlongM.E. et al.

    Autosub long range: A long range deep diving AUV for ocean monitoring

  • HobsonB. et al.

    Tethys class long range AUVs-extending the endurance of propeller-driven cruising AUVs from days to weeks

  • JohnsonN. et al.

    The importance of trust between operator and auv: Crossing the human/computer language barrier

  • JohnsonN.A. et al.

    An architecture for planner based control of autonomous vehicles

  • KwokK.S. et al.

    Analyzing the multiple-target-multiple-agent scenario using optimal assignment algorithms

    J. Intell. Robot. Syst.

    (2002)
  • LiuL. et al.

    Large-scale multi-robot task allocation via dynamic partitioning and distribution

    Auton. Robots

    (2012)
  • LysgaardJ. et al.

    A new branch-and-cut algorithm for the capacitated vehicle routing problem

    Math. Program.

    (2004)
  • M.S. Gudaitis, Multicriteria mission route planning using a parallel A* search (M.Sc. thesis), Air Force Institute of...
  • PereiraA.A. et al.

    Risk-aware path planning for autonomous underwater vehicles using predictive ocean models

    J. Field Robotics

    (2013)
  • SharmaY. et al.

    Comparison of Dijkstra’s shortest path algorithm with genetic algorithm for static and dynamic routing network

    Int. J. Electron. Comput. Sci. Eng. (ISSN: 2277-1956/V1N2)

    (2012)
  • M. ZadehS. et al.

    Optimal route planning with prioritized task scheduling for AUV missions

  • M. ZadehS. et al.

    A novel efficient task-assign route planning method for AUV guidance in a dynamic cluttered environment

  • CarstenJ. et al.

    3D field D*: improved path planning and replanning in three dimensions

  • Cited by (0)

    Somaiyeh Mahmoud Zadeh completed her Master degree in Computer Science at National University of Malaysia. She is currently Ph.D. student at School of Computer Science, Engineering and Mathematics, Flinders University of South Australia. Her area of research includes computational intelligence, autonomy and decision making, situational awareness, and motion planning of autonomous underwater vehicles.

    David M.W. Powers is Professor of Computer Science and Director of the Centre for Knowledge and Interaction Technology and has research interests in the area of Artificial Intelligence and Cognitive Science. His specific research framework takes Language, Logic and Learning as the cornerstones for a broad Cognitive Science perspective on Artificial Intelligence and its practical applications. Prof. Powers is known as a pioneer in the area of Parallel Logic Programming, Natural Language Learning, Unsupervised Learning and Evaluation of Learning, and was Founding President of ACL SIGNLL, as well as initiating the CoNLL conference. His CV includes positions at Telecom Paris, University of Tilburg, University of Kaiserslautern, Macquarie University, as well as work with industry, and commercialization of research through several start up companies. Prof. Powers also serves on several programming committees and editorial boards and being Editor-in-Chief of the Springer journal Computational Cognitive Science and the Springer book series Cognitive Science and Technology.

    Karl Sammut completed his Ph.D. at The University of Nottingham (U.K) in 1992 and was employed between 1992 and 1995 as a Postdoctoral Fellow at The Politecnico di Milano (Italy), and at Loughborough University (UK). He commenced his appointment at Flinders University in 1995. He is currently an Associate Professor with the Engineering Discipline in the School of Computer Science, Engineering and Mathematics. His areas of expertise are in embedded systems, robotics, smart structures, and maritime electronics. His particular areas of research specialization are concerned with Autonomous Underwater Vehicles (AUVs) including, navigation, optimal guidance and control systems, mission planning systems, and propulsion systems for AUVs. A/Prof Sammut is the Director of the Centre for Maritime Engineering, Control and Imaging at Flinders University and was a Cluster Project leader for the CSIRO Wealth from Oceans Flagship funded project on AUV based pipeline monitoring.

    View full text