Abstract
When unmanned underwater vehicles (UUVs) perform tasks, the marine environment situation information perceived by their sensors is insufficient and cannot be shared; moreover, the reasoning efficiency of the situation information is not high. To deal with these problems, this paper proposes an ontology-based situation awareness information expression method, using the Bayesian network method to reason about situation information. First, the situation awareness information is determined in uncertain events when performing tasks in the marine environment. The core and application ontologies of UUV situation awareness are also established. Subsequently, semantic rules are constructed, and uncertain events are identified through the corresponding semantic rules. The Jess inference engine is used to update the ontology model. Because the ontology cannot reason about uncertainty, it is transformed into the Bayesian network to reason about the impacts of uncertain events on tasks. Simulation experiments verify the effectiveness and accuracy of the situation awareness reasoning method that combines the ontology and the Bayesian network.
1. Introduction
Situation awareness was first applied in the aviation field to study pilots’ knowledge and understanding of current flight states [1]. The most famous definition of situation awareness was proposed by Endsley in 1995. That is, the situation elements in a certain time and space environment are perceived and the obtained information is understood. Then, based on this understanding, a prediction of the next moment of these situation elements can be generated. Situation awareness is particularly important for unmanned systems [2]. The characteristics of autonomy require unmanned systems to have the ability to perceive and predict environmental situations [3, 4]. In case of unmanned underwater vehicles (UUVs), the research purpose of marine environmental situation awareness is to enable the UUV to have the ability to collect and recognize various types of relevant information in the marine environment, to integrate multisensor information, to understand the state of the marine environment and the current situation of the UUV itself, to evaluate threats, and finally to guide its decision making and actions. A high level of situation awareness ability is an important prerequisite for achieving efficient decision making.
Since the 1980s, researchers have actively studied the issues of situation awareness and have elucidated several aspects including the concept of situation awareness, factors that affect situation awareness [5, 6], evaluation of situation awareness ability [7–10], team situation awareness [11–14], shared situation awareness [15], and consistency of situation awareness. These studies include qualitative analysis of situation elements, quantitative analysis and calculation of situation awareness [16–18], and the analysis of the relationship between situation awareness theory and cybernetics in the literature [19].
To deal with abnormal situations in dangerous environments, an innovative situation awareness system was proposed [20, 21], which consists of four parts: a situation information collection module, a dynamic Bayesian network evaluation module, a situation recovery module, and a human-machine interface. This research provides a graphical system for handling dynamic and uncertain information, which helps decision makers minimize dynamic risks. Considering the complex environment of naval warfare mentioned in the literature [22], a method based on the perception situation map obtained by each user and combined with the actual needs of users for perception was proposed to study the completeness, accuracy, and timeliness of the perception ability. Moreover, the final situation awareness ability was reflected by the situation awareness of all users in the system. A semantic knowledge framework was proposed in the literature [23]. This framework provided a complete system with semantic interoperability among all information sources, and it enhanced the environmental situation awareness ability of the autonomous underwater vehicle based on the expert system and the information sensed by sensors.
In the literature [24], the problem of consistent situation awareness in coordinated operations was comprehensively considered. Based on the theory of distributed situation awareness and from a system perspective, a dynamic model of the team’s consensus situation awareness process was constructed. Taking the shortest time required for internal and external consistency of the system as the final indicator, a measurement method for team situation awareness consistency was provided. In the literature [25], multilayer semantic information integration was conducted according to ontology, semantic web technology, and rule-based reasoning. A military scene ontology and combat management ontology were defined as the domain ontology. Moreover, a situation awareness ontology was proposed as the core ontology of the integrated military scene and combat management ontologies. Through autonomous information fusion and reasoning, situational awareness was improved. A situation awareness support system was developed as described in the literature [26] to deal with uncertain situations. A multiperspective method was applied to systematically evaluate the degree of situation awareness in the situation awareness support system. This method included two situation awareness indicators: one for the objective and subjective measurement of situation awareness and the other for the evaluation of the operator’s workload. The experimental results showed that the situation awareness support system improved the situation awareness ability of the operator.
There are also shortcomings in improving situation awareness ability, such as the lack of knowledge expression methods for situation awareness information and insufficient operability at the semantic level. The expression of situation awareness information is the basis of method research on situation awareness. There are various approaches for expressing information, including predicate logic notation, semantic web notation, frame notation, and object-oriented notation. The common disadvantage of these methods is the lack of knowledge sharing and reusability. Data mining and some improved algorithms are applied to situation awareness data integration, making situation information highly integrated, which lays the foundation for the next situation assessment. The solutions to the problems such as a qualitative and quantitative analysis of situation awareness and model consistency often adopt adaptive estimation and fuzzy theory.
The rest of this paper is organized as follows: (1) ontology-based marine environment situation awareness information modeling, (2) triggering of UUV uncertain events based on SWRL rules and Jess, (3) reasoning the impact level of uncertain events based on Bayesian network, and (4) simulation verification and result analysis.
2. Ontology-Based Marine Environment Situation Awareness Information Modeling
2.1. Detection of Perceived Information in Uncertain Events
When performing tasks in a complex underwater environment, it is difficult for UUVs to accurately predict and describe the marine environment by constructing a mathematical model. In such an environment, some uncertain events may occur, such as target threat, abnormal thruster, abnormal steering gear, insufficient energy, and water leakage events. The uncertainty of these events is reflected in the time when the event occurs and whether they can happen or not. The UUV detects and analyzes the corresponding perceived information on these uncertain events to provide the threat level of the uncertain events, which lays the foundation for the next situation assessment. Uncertain events can be divided into three categories: marine environment, platform status, and task status. The UUV detection data according to the types of uncertain events are shown in Table 1.
2.2. Construction of UUV Marine Situation Awareness Information Ontology Model
To realize the sharing and reusability of knowledge in the field of UUV marine situation awareness, an ontology is used to model the situation awareness information in the UUV complex marine environment. The UUV core ontology describes the most basic concepts in the UUV field and the relationship between these concepts, including the most basic components and functions of the UUV, as shown in Figure 1.
The concepts involved in the basic components and functions of the UUV shown in Figure 1 are as follows:(1)Platform: UUV.(2)Payload: including sensors, hardware, and instruments on the UUV platform.(3)Modules: software that constitutes the UUV.(4)Sensors: devices that can perceive information from the external environment, the UUV’s own pose status, and coordinate communication.(5)Drive: a software module that serves the hardware.
A UUV situation awareness application ontology is constructed on the basis of the UUV core ontology. Three corresponding application ontologies are designed according to three types of uncertain events: the marine environment application ontology, UUV platform status application ontology, and task status application ontology.
2.2.1. Marine Environment Application Ontology Modeling
Create the classes and attributes in the marine environment application ontology, as shown in Table 2. The structure of the marine environment application ontology is shown in Figure 2.
2.2.2. UUV Platform Status Application Ontology Modeling
Create the classes and attributes in the UUV platform status application ontology, as shown in Table 3. The structure of the UUV platform status application ontology is shown in Figure 3.
2.2.3. UUV Task Status Application Ontology Modeling
Create the classes and attributes in the UUV task status application ontology, as shown in Table 4. The structure of the UUV task status application ontology is shown in Figure 4.
After modeling the application ontologies, values need to be assigned to them. Combined with domain knowledge, attribute assignment is performed on various instances. For example, the attribute of the threat degree of the marine environment to the UUV can be assigned the attributes safe and dangerous. After instantiation, the domain knowledge base of the UUV can be formed.
3. Triggering of UUV Uncertain Events Based on SWRL Rules and Jess
As a language for describing inference rules, SWRL has two main forms: XML and RDF. The structure of SWRL is divided into four parts: Imp, Atom, Variable, and Building. Imp is responsible for the composition of SWRL rules, consisting of head and body. Head usually describes the conclusions obtained by reasoning, while body describes the conditions required for reasoning. Therefore, the rule expression is condition (body) ⟶ conclusion (head). Atom is composed of constraint expressions; this paper mainly uses two definitions, namely, C (? x) and P (? X, ? y), in Atom. Variable represents all variables used in Atom. Building is a modular component of SWRL, which is used to define logical comparison relationships.
In this paper, based on expert domain knowledge, SWRL rules for UUV uncertain events are designed. For example, Target_Distance (?x)∧hasDistance (?x, ?y)∧swrlb: lessThan (?y, 130) ⟶ Marine_Environment (?x). The part before the reasoning symbol “⟶” is the body part of the reasoning condition in the Imp of the SWRL rule description part. The part after the symbol is the head part of the reasoning result. Target_Distance (?x), hasDistance (?x, ?y), swrlb:lessThan (?y, 130), and Marine_Environment (?x) are the basic components of the head and body. Target_Distance (?x) and Marine_Environment (?x) are concepts defined in the marine environment ontology. HasDistance (?x, ?y) is the attribute defined in the marine environment ontology. swrlb: lessThan (?y, 130) is the numerical comparison relationship represented by Building. ?x and ?y represent the variables used. This rule implies that when the distance between the UUV and the target is less than 130 m, the target threat event is currently triggered.
The Jess inference engine was developed by Ernest Friedman-Hill of Sandia Lab in the United States. It is the fastest among general rule inference engines. The Jess inference engine is used to reason about the ontology knowledge base by applying declarative rules. However, the reasoning rules described by SWRL cannot be directly used for reasoning by Jess but must be converted into the grammatical format required by the Jess inference engine. The UUV situational awareness uncertain event ontology and SWRL rules are loaded into the Jess inference engine. Then, the fact base and the rule base are converted to perform inference in the inference engine. The fact and rule bases of the ontology and SWRL rule conversion inference engine are shown in Figure 5. Jess is run for reasoning, and the obtained reasoning results are imported into the UUV situation awareness uncertain event ontology to update the ontology model.
4. Reasoning the Impact Level of Uncertain Events Based on Bayesian Network
The ontology has an absolute advantage in describing deterministic knowledge. There are a number of uncertain factors in the complex marine environment, leading to the generation of uncertain knowledge. The uncertain knowledge provides important data support for UUV system decision making. However, the ontology cannot express and reason the uncertain knowledge. Probability expansion of the OWL language of the ontology needs to be conducted. The state class, probability class, variable class, conditional probability class, and prior probability class are set in OWL; here, the probability class is the parent class of the conditional probability class and prior probability class.
Through the application of the BayesOWL method, the probabilistically extended OWL ontology is converted into a Bayesian network. The flowchart of the conversion of the situation awareness ontology into the Bayesian network is shown in Figure 6. The structure diagram of the Bayesian network constructed according to the rules is shown in Figure 7. The conditional probability table is shown in Table 5. The Bayesian network reasoning is used to comprehensively analyze the perception information, infer the threat level of the current information to the task, and finally instantiate it into the UUV situation awareness ontology for UUV re-planning.
5. Simulation Verification and Result Analysis
A UUV performs survey tasks in a certain sea area, during which uncertain events may occur, such as appearance of random obstacles, insufficient energy reserve, sensor failure, abnormal task load, and other situation awareness events. Uncertain events are identified according to SWRL rules, and on the basis of the Bayesian network, the impact of uncertain events can be inferred. Then, the UUV can judge whether it is necessary to re-plan in order to complete the task. In this study, the events of the UUV encountering random obstacles and insufficient energy are simulated to verify the effectiveness of the proposed UUV situation awareness method based on the ontology and Bayesian network.
5.1. Simulation Analysis of Triggering Obstacle Events
In the process of UUV performing tasks, the forward-looking sonar sensor carried by the UUV detects the target in front of it. The information of the target detected by the sensor is represented in the form of examples in the UUV’s situation awareness uncertain event ontology model. The sailing speed of the UUV is 2 m/s. Assume that when t = 200 s, the UUV perceives that the distance between the target and the UUV is 140 m. According to expert experience, as the distance from the target decreases and becomes less than 130 m (which is the safety threshold), a target threat event is considered to have occurred. Then, according to the preset SWRL rules, it is determined that the target threat event is generated; the target instance is generated simultaneously and stored in the OWL ontology model file. The rules involved are as follows: Target_Distance (?x)∧hasDistance (?x, ?y)∧swrlb: lessThan (?y, 130) ⟶ Marine_Environment (?x).
The node evidence in the BN of the uncertain event in Figure 7 is updated in real time. According to BN inference, the impact level of the target threat event on the task is shown in Table 6, and the corresponding curve is shown in Figure 8.
As shown in Table 6, when the task is executed for 220 s, the threat probability of the obstacle to the task is 0.66. As the distance to the obstacle decreases, the threat of the obstacle to the task increases. At 225 s, the threat probability of the obstacle to the task is 0.83, which is greater than the critical threshold (0.8), thus triggering the target threat event. At this time, the UUV calls the collision avoidance program for obstacle avoidance processing, and the UUV platform adjusts the speed and direction in time. Subsequently, the threat probability of obstacles to the task gradually decreases. It can be seen that the ontology and BN are highly effective for situation awareness information processing in the process of UUV performing tasks.
5.2. Simulation Analysis of Triggering Insufficient Energy Margin Events
In the process of UUV performing tasks, due to the limited energy carried by the platform, energy shortage events may occur. The energy margin is judged according to the battery voltage detected by the sensor on the UUV. The information of the energy margin is expressed in the form of examples in the UUV’s situation awareness uncertain event ontology model. Assume that the sailing speed of the UUV is 2 m/s. According to expert experience, after the UUV works for a period of time, when the battery voltage of the instrument is less than 3.8 V, the energy margin is considered to have an impact on the UUV task. Then, according to the preset SWRL rules, it is determined that the energy shortage threat event is generated; the instance of insufficient energy margin is generated simultaneously and stored in the OWL ontology model file. The rules involved are as follows: Power_Voltage (?x)∧hasVoltage (?x, ?y)∧swrlb:lessThan (?y, 3.8) ⟶ Energy_Allowance (?x).
The node evidence in the BN of the uncertain event in Figure 7 is updated in real time. According to BN inference, the task impact level of the energy margin on the UUV is shown in Table 7 and the corresponding curve is shown in Figure 9.
As shown in Table 7, when the survey task is executed for a period of time, the probability of detecting the impact of the energy margin on the task at time t is 0.45. With the task over time, the impact level of the energy margin on the task further increases. At t + 4, the probability of the impact level of the energy margin on the task reaches 0.83 (the battery voltage of the instrument is less than 3.8 V). At this time, an instance of the insufficient energy margin event occurs. Then, the UUV platform terminates the task and the UUV returns to the recovery point. In the process of returning, because the energy margin is decreasing, the probability of the impact of the energy margin on the task increases until it reaches 1. It can be seen that the combination of the ontology and BN is highly effective for situation awareness information processing in the process of UUV performing tasks.
6. Conclusions
Given the growing national significance of marine engineering and marine research, the development of UUVs has been attracting increasing attention. The UUV’s perception of situation information is particularly important when performing tasks in a complex marine environment. In this study, through the detection of data related to the marine environment, platform status, and task status and by using the ontology to express situation awareness information, the UUV situation awareness core and application ontologies were constructed. SWRL rules were used to identify uncertain events. Finally, the ontology was transformed to the Bayesian network to reason about the impact of uncertain events. Simulation experiments verified the effectiveness of the UUV situation awareness method based on the ontology and Bayesian network. The proposed situation awareness method can also be used in other unmanned systems. In the future, high-efficiency and real-time issues of UUV situation awareness will be further studied.
Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
This research was supported by the Basic Scientific Research Project of Heilongjiang Province Department of Education (no. 135309487).