Automated context aware composition of Advanced Telecom Services for environmental early warnings
Introduction
An Advanced Telecom Service is defined as a service that exploits the convergence of communication networks and takes also advantage of features accessible from the World Wide Web (Object Management Group [OMG], 2012). This kind of composite services are needed when the request of a client cannot be satisfied by a single service. Traditionally, the composition of Telecom Services in the industry is carried out through interactive graphical interfaces called Services Creation Environments (SCE),1 which allow the user to articulate services functionalities using drag-and-drop tools. This approach is usually valid in Telecommunications thanks to the fact that the composition is performed using exclusively Telecom services, whose reliability is very high (99.999%) (Chrighton, Long, & Page, 2007). However, composing Advanced Telecom Services including services from the Internet, commonly called Web Services, is a very different matter. Due to the dynamic nature of the Internet, Web Services may change, become unavailable and grow in number to unmanageable sizes. This means that composition of advance telecom services is unfeasible in practice if the user employs traditional methods.
Previous works from both the academia (Hatzi et al., 2012; Hoffmann, Bertoli, & Pistore, 2007) and industry2 have revolved around this topic. However, few academic approaches include implementations in real world scenarios, and few works from the industry are publicly open and easy to extrapolate to similar cases, as exposed by Dustdar and Schreiner (2005) Furthermore, previous work in the area focuses on representing the services using semantic descriptors (Huang, Lee, & Crespi, 2012) and applying afterwards different techniques, such as automated planning (Oh, Lee, & Kumara, 2006) or genetic algorithms (Ye, Zhou, & Bouguettaya, 2011). Additionally, some of these approaches have obtained promising results using Web Services, but no prior work that we are aware of is able to combine them with Telecom features. Another important limitation of these approaches is that they require that the users express their requests using formal languages such as Golog (Sohrabi, Prokoshyna, & McIlraith, 2009) or PDDL (Hatzi et al., 2012). This means that the interaction of a large number of people with little or no technical background with the system is unfeasible, which renders these works useful only in a limited set of scenarios.
The goal of the environmental early warning field is to create contingency plans that help resolve potentially harmful or dangerous situations, such as floods or tropical storms. The detection of these situations combines both information gathered from sensors around the monitored area and the input of relevant users. An example of such a situation would be evacuating all the villages close to a river after detecting that its level has risen beyond regular measurements or upon the request of an observer. In this case a contingency plan would include actions to monitor the river, determine affected areas, warn the villagers and coordinate the logistics of the evacuation. Since the participation of a human is required in critical scenarios, we assume that all the requests are triggered by users. Besides, in rural environments the access to a device able to send a request in the format specified by the system may be limited. This means that the system should be able to process natural language so it can be initiated through simple communication means like a phone call or an SMS. In fact, thanks to the rather restrictive process of the composition of services in this domain, the main procedures and their associated services can be described using simple semantic annotations by experts in the field, which allow a natural language recognition technique to be implemented without imposing significant restrictions to the potential users. This also means that an automatic translation of the input into a formal language without intervention from either the user or an expert is a much easier task.
Recent studies have explored the application of Natural Language processing techniques to the Automated Composition of Services field, especially in intelligent home environments (Cremene et al., 2009) These approaches offer mechanisms to map user words to basic functionalities of the system. In those works, the goal is to extract the workflow of the composite service from the User Request. The novelty of the application of Natural Language in this work is that in some scenarios, such as the early environmental warnings domain, the user can only express his desire or goal and not the workflow of the composite service. The workflow of the new composite service should be obtained from the automated composition process. Another element of particular relevance is the user preferences and the context of the situation. Automatically discerning the context of the request can enrich the process of automated composition processing by adding important information. In this regard, special words in user requests such as “urgently”, “danger” and “quickly” can represent a preference for service quality instead of cost. Secondly, data about the user location and the characteristics of the device may offer critical information about the preferred delivery format for the output of the composition of services.
In relation to the usefulness of automated services composition in this work, some considerations must be made. The Environmental management domain contains the following particularities: (i) composed services can be formed by Web Services and/or by basic Telecom features like Call or send SMS, (ii) the number of different services is rather limited due to the specialized field of action, (iii) in environmental management the procedures for emergencies handling and monitoring are standard. This means that these procedures and the associated services can be described using semantic annotations by experts, as the number of annotations needed is relatively few and standard. This is in fact a very important issue, since few service providers have taken up the opportunity to mark-up their web services, for the simple reason that they do not envisage the use of their services by an automated planning system (Carman, Serafini, & Traverso, 2003). Although the present architecture is focused on a particular domain, that is, the implementation is domain-dependent, most of the principles discussed here are applicable to similar scenarios.
The overall functioning of the architecture is as follows: the user request is received in natural language from a given device. The request is processed to determine the goals and preferences of the user. At the same time, information obtained from the sensors may be added to the request depending on the context. Next, the request is translated dynamically into a planning instance modeled using the Planning Domain Definition Language (PDDL) (Gerevini, Haslum, Long, Saetti, & Dimopoulos, 2009). Then, the PDDL formatted request is sent to the High Level Replanner module of the Planning, Learning and Execution Architecture (PELEA) (Guzmán et al., 2012). PELEA computes a plan using a domain-independent planner and manages its execution (note that the plan represents the composition of services). Finally, the composed plan is executed in a Jain SLEE 5 environment for convergent services.
In this context, the main contributions of our research work are: (1) A Framework for Advanced Telecom Services Composition based on Automated Planning; (2) a metamodel for user context including device characteristics, preferences and user profile information; (3) a mechanism that automatically translates user requests in Natural language into planning instances formulated in PDDL; and (4) the integration of a planning architecture in the execution of plans based on expert-made Mashups.
This paper is organized as follows: Section 2 presents the motivating scenario of early warning in environmental domain. As the AUTO framework is based in Automated Planning; In Section 3 the whole architecture of the framework is described. Section 4 exposes the background about planning and describes the planning domain construction. Section 5 describes the modeling of user request and his translation to PDDL. Section 6 Describes the Prototype and experimentation. Section 7 presents the related work and Section 8 draws the conclusions.
Section snippets
Environmental management domain
A sketch of an environmental management system is presented in Fig. 1. The role of the environmental manager is to make decisions about the environmental alarms and crop management of a region. For this purpose he/she can request information from the network of sensors deployed at several spots. The environmental manager can also use Telecommunication and Web services to process basic data and send information to both farmers and actuators. Available services often change dynamically and the
Architecture of the AUTO framework
The architecture of AUTO is depicted in Fig. 2. The modules may be deployed in different machines so the computational load of the different processes can be distributed. The access method can be either voice or text, which means that AUTO can be accessed from a broad range of devices. In the literature, other alternatives have been proposed for user request treatment, such as Mashups (Zhao, Bhattarai, Liu, & Crespi, 2011) or service creation environments (Laga, Bertin, Glitho, & Crespi, 2012).
Automated planning and specification of the domain
Automated Planning is the task of finding a sequence of actions (plan) that leads to a state where the goals stated in the definition of the problem are achieved. The goal state is obtained by executing the actions of the plan from a given initial state. A domain independent planner takes two inputs: a domain description and a problem description. Both are usually described using the standard language PDDL. The planning domain can be understood as a set of statements defining the types of the
Request analysis module
The Natural language analysis module receives as input the user request done from his/her mobile device in NL, and generates a planning instance in PDDL that represents the state and the goals of the user. The Request Analysis Module is composed of two sub-modules: the sub-module NL Analysis, which parses and processes the user request in natural language; and the sub-module Context Analysis, which takes into account user preferences, device capabilities, and situation (either from some special
Prototype and experimentation
In this section a more detailed description of the implementation is given. Also in order to assess the viability of AUTO some experimentation regarding both performance and accuracy will be done. As the AUTO framework is composed of different modules, the experimentation was conducted for each module by separate. Here, we present the results for User Request Analysis and the Automated Planning. The software associated with AUTO Framework including the remaining modules source and binaries are
Related work
Zhu, Zhang, Chen, and Cheng (2010) presented Hybrid Service Creation and Execution Environment (HSCEE), a Template-based service creation platform with low latency service execution. HSCEE is based in on BPEL templates but most of the design tasks must be done manually. Natural language processing is also not considered.
The OPUCE Project8 (Yu, Sheng, Han, Wu, & Liu, 2012) (Yelmo, del Álamo, Trapero, & Martín, 2011) presented a User-centric service creation environment (SCE)
Conclusion
In this paper we extend AUTO by integrating Automated Planning into an existing architecture as the deliberative process. This is a more general approach than the previously used technique that allows a greater flexibility both in terms of design and the implementation of the underlying algorithms. The main contribution is the automatic generation of planning instances in PDDL from natural language requests. Furthermore, we allowed the use of preferences modeling them as the metrics of the
References (30)
- et al.
Towards runtime discovery, selection and composition of semantic services
Computer Communications
(2011) - et al.
Deterministic planning in the fifth international planning competition: PDDL3 and experimental evaluation of the planners
Artificial Intelligence
(2009) - et al.
Constructing composite web services from natural language requests
Web Semantics: Science, Services and Agents on the World Wide Web
(2010) - et al.
A user-centric approach to service creation and delivery over next generation networks
Computer Communications
(2011) - et al.
A semantically enhanced service repository for user-centric service discovery and management
Data & Knowledge Engineering
(2012) - Carman, M., Serafini, L., & Traverso, P., (2003, June). Web service composition as planning. In ICAPS 2003 workshop on...
- et al.
Workpad: process management and geo-collaboration help disaster response
International Journal of Information Systems for Crisis Response and Management (IJISCRAM)
(2011) - Chrighton, C., Long, D. T., & Page, D. C., 2007. JAIN SLEE vs SIP Servlet Which is the best choice for an IMS...
- et al.
Natural language processing (almost) from scratch
The Journal of Machine Learning Research
(2011) - et al.
Service composition based on natural language requests
A survey on web services composition
International Journal of Web and Grid Services
A look-ahead B&B search for cost-based planning
Introduction to emergency management
An integrated approach to automated semantic web service composition through planning
IEEE Transactions on Services Computing
Cited by (16)
Data quality challenges in large-scale cyber-physical systems: A systematic review
2022, Information SystemsCitation Excerpt :CPSs are an active area of research [46,47], with a significant importance to the future of smart cities [48]. CPSs are multidisciplinary cross-domain information systems which bring together different sectors of smart cities’ public services, such as smart transportation management, smart utility management, smart buildings [49], smart environment management [50] and smart governance, Where data sensing, knowledge extraction, and higher automation are critical elements in the future of these services [51]. The future cities can provide smarter services by utilising IoT solutions that relay on an extended number of sensors and can provide scalable and interactive functionalities, e.g., the city scale traffic management solution illustrated in Fig. 1, [52].
FESC: Functionally Equivalent Service Composition
2020, Internet of Things (Netherlands)Citation Excerpt :In [34], a six phase composition process is suggested to provide semantic context-aware services composition for building home automation system. A framework for composition approaches based on automated planning in Telecommunications scenarios is suggested in [4]. However, monitoring the environment is not discussed in [4].
Rich dynamic mashments: An approach for network management based on mashups and situation management
2016, Computer NetworksCitation Excerpt :Considering, first, the above results. Second, the mechanism for generating RDMs has similar time-composition behavior than the composition proposals introduced on other application domains [67]. We can state that, in terms of time-composition, it is feasible to use the proposed approach to coping effectively with nmsits.
Comparing Drools and Ontology Reasoning Approaches for Automated Monitoring in Telecommunication Processes
2016, Procedia Computer ScienceTowards automated composition of convergent services: A survey
2015, Computer CommunicationsCitation Excerpt :This high number of services turns complex the use of traditional methods, such as SCEs and MDEs, for creating convergent services. Therefore, some of the research approaches analysed here especially the group of approaches based on automated planning may be used for discovering convergent services in Telecom infrastructure; for example, Ordóñez et al. [109] present a framework for including automated composition approaches based on automated planning in Telecommunications scenarios. On the other hand, recent algorithms such as [147,148], a uses mechanisms for reducing the services search space by pruning candidates by dominance relationships and constraint validations at candidate level.