Seminar on Intelligent Personal Assistant based on Internet of Things approach

Seminar on Intelligent Personal Assistant based on Internet of Things approach

• 1.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 1 CHAPTER 1 INTRODUCTION The evolution of the Internet into the global communications network, omnipresent and universal, has supported the development of new paradigms and technologies. Moreover, the advances verified in areas such as speech recognition, natural language processing, semantic web, machine learning, and artificial intelligence, combined with the huge amount of available information made accessible by the Internet, has enabled the creation of intelligent personal assistants (IPAs). IPAs are mobile, autonomous, and software agents capable to perform tasks or services on behalf of humans. The services provided by IPAs are not optimized often, since access to databases and other resources are limited many times. Then, the learning mechanism of IPAs could be improved by the integration of information generated from the surrounding environment, such as the temperature, humidity and luminosity, among others. These external indicators could be measured with sensors that would then transmit the gathered values to the IPA. The values can be used by IPAs to generate useful information to trigger services and assist people on their daily life activities. This process can be seen as natural interaction among people, environment, and machines, creating a scenario of ubiquitous computing. The Internet of Things (IoT) is a relatively new paradigm that uses this ubiquitous computing concept. The vision of IoT can be seen as the evolution of the Internet, where almost every object can support an IP address and consequently be connected to the Internet. Personal assistants are not recent, being used in daily business and personal tasks for many years. Personal assistants are responsible for daily task management such as scheduling meetings, reserving hotel rooms, shopping, and paying bills, among others. In the past, the personal assistant role was exclusively performed by people but, nowadays, it can be performed by a digital device with learning capabilities .Generally ,the personal assistants available on these digital devices are able to make calls, send messages, use the calendar, open web pages, read news, and get weather data, among other options. In the 1990s, the IPA concept was introduced. The use of techniques from the AI field created the possibility of constructing intelligent machines that autonomously performed tasks on a user’s behalf. The use of mechanisms such as data mining and machine learning algorithms was important in creating more responsive and self-aware machines.
• 2.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 2 The main contributions of the paper are the following: 1) Over viewing IoT technology and reviewing the available enabling technologies and protocols. 2) Showing the main features of IPAs providing some examples of approaches involving these features. 3) Describing how IoT networks may improve IPAs functionalities.
• 3.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 3 CHAPTER 2 INTERNET OF THINGS The internet of things is the internetworking of physical devices, vehicles, buildings and other things, embedded with electronics, software, sensors, actuators and network connectivity that enables these objects to collect and exchange data. IoT vision “brings life” to almost every device by assigning an identification and an IP address to almost every object that can generate, pass on, or receive some kind of information. Sensors integrated into objects can gather data from environment that can be processed later. A wireless sensor network (WSN) can include a great number of nodes (dozens, hundreds, or even millions). The ability to connect a WSN to the Internet provides a great opportunity to deliver information in real time. There are two main approaches to provide IP addresses for sensors and connect the network to the Internet: 1) a proxy- based and 2) a sensor node IP stack-based approaches [7]. The proxy-based approach uses a sink node that serves as a proxy, providing a bridging mechanism between a WSN and the Internet. Sensor nodes use dedicated protocols that interact with a sink node, which also supports the TCP/IP, enabling Inter- net communication. The main disadvantage of this approach is the gateway functionality, as it involves a complex development process and can be identified as a point of failure. In contrast, the sensor node IP stack-based approach relies on IP as the routing protocol inside a WSN, which allows direct communication among the nodes and the Internet. A gateway can also be used to connect the WSN to the Internet, filtering all the unwanted traffic. With the creation of IoT networks, the amount of data generated has increased dramatically because the number of devices generating traffic inside the network is much larger than before. That data can enable the creation of new types of services and applications. IoT can have a big impact on businesses by automating some processes and improving the control of many environment variables, which eases the decision-making process.
• 4.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 4 Despite all the recent projects using IoT technology, a standard approach is not widely followed. An initial conceptual architecture has been accepted by the scientific community based on simple context factors. This architecture is composed of three layers: the perception layer, network layer, and application layer, as depicted in Fig1. To turn IoT into a feasible technology, many standardization efforts have been performed. A standard for the physical and medium access control (MAC) layers, the IEEE 802.15.4, has been approved. In addition, an overlay protocol was created to enable the fragmentation of the IPv6 packets inside IEEE 802.15.4 frames, called IPv6 over low-power wireless personal area networks (6LoWPAN). A routing protocol was defined for low-power and lossy networks (LLNs), referred to as the ripple routing protocol (RPL). A web protocol was also created for LLNs, called the constrained application protocol (CoAP). Besides CoAP, other protocols are being considered to serve as the standard IoT application layer protocol, such as the extensible messaging and presence protocol (XMPP) and the message queue telemetry transport (MQTT). 2.1. IEEE 802.15.4 Using power-constrained objects (such as sensors) on IoT networks can cause various performance issues. To attenuate these issues, the Institute of Electrical and Electronics Engineering (IEEE) created in 2003 the standard IEEE 802.15.4, which defines the physical (PHY) and MAC layers of low-rate wireless personal area networks [8], [31]. It has a maximum data rate of 250 kb/s, much lower than the rate defined in the IEEE 802.11 standard. This fact allows small and power-constrained nodes to consume less energy. The standard IEEE 802.15.4 has a frame sizeupto127B,with two transmission modes ,beacon- enabled mode and non-beacon-enabled mode ,in the MAC layer. Devices using the IEEE 802.15.4 can communicate over areas covering 100 m on single-hop architectures. Transmissions beyond this range require the cooperation of neighbor nodes to deliver the message to its final destination using a multihop architecture. Two types of nodes can exist on a multihop architecture: a full- function device (FFD) and a reduced-function device (RFD). An FFD can serve as the coordinator of a personal area network (PAN) and communicate with other nodes in the network. RFD has a lower processing capability, so it Figure 2.1. Illustration of IoT conceptual architecture
• 5.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 5 can never act as the PAN coordinator, communicating only with FFD nodes. There are two main network topologies, the star topology and the peer-to- peer topology, as depicted in Fig2. 2.2. 6LoWPAN 6LoWPAN, developed by the Internet Engineering Task Force (IETF),is an essential protocol for all IoT IP-based architectures using the standard IEEE 802.15.4, Using IPv6 as the Internet protocol in IP-based IoT architectures allows the presence of more devices on the network (2128 ) compared to IPv4 (232).It also uses the stateless address auto configuration mechanism to assign automatically an IPv6 address to a device at the moment of its initialization. However, it is impossible to encapsulate directly IPv6 packets (1280-b MTU) into IEEE 802.15.4 frames. To solve this problem, the 6LoWPAN protocol was created; it compresses the size of packets flowing in the network, decreasing both the bandwidth and the energy consumed by power-constrained devices. Beyond adapting the data packets’ size, 6LoWPAN is capable of resolving addresses, implement addressing management mechanisms, and discover devices and services. Various open- source operating systems that deploy the 6LoWPAN stack are available, such as Contiki OS with its uIPv6stack or Tiny OS with its BLIP (Berkeley low-power IP stack). 2.3. Ripple Routing Protocol – RPL In2008, the IETF Working Group on routing over low-power and lossy networks (ROLL) was created. The ROLL Working Group intended to create a routing protocol for LLNs that would support a wide variety of link layers with common requirements such as low bandwidth, lossiness, and low power. The protocol was named RPL [37]–[39]. The RPL is an IPv6 distance vector routing protocol oriented for directed acyclic graph topologies, in which no direct cycles connect the network devices. Generally, such networks have a root Figure 2.2. Illustration of star and peer-to-peer topology in IEEE802.15.4
• 6.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 6 node or a set of root nodes to coordinate tasks and collect data. For each root node, a destination-oriented directed acyclic graph is created using an objective function that defines how the routing metric is computed and ranks to encode the distance of each network node to its reference root. The RPL can be deployed on different operating systems, such as Contiki OS and TinyOS. Ancillotti addressed the interplay between RPL and different address auto configuration algorithms, using the Contiki RPL to conduct the practical experiments. 2.4. Constrained Application Protocol – CoAP IoT devices can be integrated on the web, creating the “Web of Things” concept. This allows Representational State Transfer (REST) architectures to be used on IoT applications. Many of those REST architectures use HTTP, which may create performance issues when used on LLNs due to its large overhead. To overcome these issues, the CoAP was created in 2010. CoAP is based on the REST architecture and uses the same methods available on HTTP. Despite that, CoAP is not a substitute for HTTP, being easily interoperable through proxy mechanisms. As shown in Fig. 3, the main difference is the use of UDP as the transport layer protocol on CoAP and the TCP on HTTP. The application layer on CoAP is divided into two sub layers: the message and request/response layers. The message layer controls message exchanges over UDP between two endpoints. Messages are specified by IDs to detect duplicates, assuming one of four different types: confirmable, non-confirmable, acknowledgment, and reset. The request/response layer is responsible for the transmission of requests and responses used to manipulate resources. A REST request is piggybacked on a confirmable or non-confirmable message, while a REST response is related to an acknowledgment message. When Figure 2.3. Comparison of CoAP and HTTP protocol stacks
• 7.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 7 confirmable messages are used, acknowledgment messages are required to confirm the receipt of the desired information. If a response does not arrive at the client (or the client receives a reset message), the message is retransmitted. With this mechanism, CoAP ensures communication reliability without using TCP, operating also as a QoS mechanism. According to the literature, various CoAP deployments on IoT projects have been implemented. It was concluded that CoAP has a better response time for client requests than HTTP. Eliasson proposed an IoT solution for the mining industry based on rock bolts and CoAP-enabled sensors to raise the security inside mines. 2.5. Extensible Message and Presence Protocol - XMPP Standardized by the IETF, the XMPP is based on the extensible markup language, which allows two or more entities to communicate inside a network. XMPP is particularly suitable for real-time communication such as instant messaging. XMPP is a scalable protocol that allows the specification of extension protocols, designated by XMPP extension protocols (XEPs). XEPs allow the addition of new functionalities to the XMPP protocol, introducing the ability to adapt to specific environments. Some XEPs can be useful for the IoT domain, including XEP-0323, XEP-0324, XEP-0325, and XEP-0326. The authors presented a lightweight implementation of the XMPP for the Contiki OS, the μXMPP. The μXMPP is configured to follow LLN specification, using only essential XMPP characteristics to reduce memory usage on the devices. Klauck and MKirsche showed an improvement to the μXMPP solution with the addition of new modules to the μXMPP architecture, enabling IPv6 support, short JIDs, temporary subscription for presence, and support for new XEPs, such as XEP- 0045 and XEP-0174. 2.6. Message Queue Telemetry Transport - MQTT MQTT, developed by the international business machines, is a lightweight application layer protocol based on a publish/subscribe method. With MQTT, clients can subscribe to information (data types) of interest by registering with a broker using TCP connections. Each message data type is referred to as a topic, and clients can subscribe to multiple topics. Every message sent on a topic is received by the clients that subscribed it. The MQTT protocol is widely used in M2M communication, and it can be used in IoT environments. It also supports three different QoS levels, which allows the traffic inside the network to be differentiated. However, MQTT uses TCP as the transport layer protocol, provoking some inefficiency
• 8.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 8 when used on power-constrained devices. To better adapt the MQTT protocol to LoWPANs, MQTT-S (MQTT for sensor networks) was created; it defines a UDP mapping for the MQTT, adding broker support for indexing topic names at the same time. The authors discussed the importance of the publish/subscribe method for WSNs and applications, where the content itself is more important than its origin. To test the MQTT-S architecture (consisting of a MQTT- S client, a message broker, and a MQTT-S gateway), the authors used the IBM wireless sensor networking test bed with two types of clients: ZigBee clients and TinyOS clients.
• 9.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 9 CHAPTER 3 INTELLIGENT PERSONAL ASSISTANTS Personal assistants are not recent, being used in daily business and personal tasks for many years. Personal assistants are responsible for daily task management such as scheduling meetings, reserving hotel rooms, shopping, and paying bills, among others. In the past, the personal assistant role was exclusively performed by people but, nowadays, it can be performed by a digital device with learning capabilities. Generally, the personal assistants available on these digital devices are able to make calls, send messages, use the calendar, open web pages, read news, and get weather data, among other options. In the 1990s, the IPA concept was introduced. The use of techniques from the AI field created the possibility of constructing intelligent machines that autonomously performed tasks on a user’s behalf. The use of mechanisms such as data mining and machine learning algorithms was important in creating more responsive and self-aware machines. Machine learning algorithms can be divided into two different categories: supervised learning and unsupervised learning methods. IPAs can normally interact equally with other intelligent objects in the environment (human or machine) to obtain knowledge about different domains. These scenarios are referred to as MASs. MASs are composed of multiple heterogeneous interactive intelligent agents within an environment, enabling parallel processing inside the system and making it less prone to failures. In MASs, agents in the environment can be influenced by other agents or by the environment itself. In single-agent systems, only the environments can influence the behavior of the intelligent agents. Intelligent agents in MAS scenarios can gain awareness of what they must do under the influence of other agents in the environment, the environment itself, and their own prior knowledge about multiple subjects. It is similar to having a network of self-conscious machines that can make their own decisions in such a way that seems natural by human standards. This fact may be useful in the construction of optimized IPAs. Some smart phones provide a personal digital assistant as one of their main features. Apple’s Siri, Google Now, Samsung’s S Voice , LG’s Voice Mate , and Microsoft’s Cortana are examples. Generally, these digital assistants make use of natural language user interfaces (NLUIs) to interact with users. NLUIs allow human– machine interaction through the translation of voice commands into machine-level commands, an important speech
• 10.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 10 technology for human-to-machine (H2M) communication. H2M communication is a field of great interest in the scientific community. Besides the already mentioned Siri, Google Now, Samsung S Voice, LG Voice Mate, and Microsoft Cortana, which are used to fulfill minor daily life issues, there are multiple projects involving IPAs and similar devices. For instance, Pozna discussed available issues in a personal IPA name Field Description Siri, Google Now, S Voice, LG Voice Mate,Microsoft Cortana General Personalassistant used to fulfil minor daily life activities. Generally used for personal interest Intelligent assistant General Desktop intelligent assistant built to help the users with communication, information access,and time management. iAPERAS Sports Personalassistant useful for nonprofessional athletes obtains dedicated information about their own health indicators. Health Pal Healthcare Personalassistant used on ambient- assisted living environments. BeWell Healthcare Unobtrusive personal assistant used to monitor the sleeping patterns of their users. Student Life Healthcare Personalassistant used to evaluate the mental well-being of college students during the school semester. PPCare Healthcare Personalassistant useful to monitor the daily life and the physiological indexes of elderly people. MAS-base assistant Manufacturing Personalassistant framework for manufacturing business based on MAS and CBR techniques. Adele Education Desktop personal assistant developed to help students learn new subject Intelligent network-based assistant PersonalCommunication Context-aware personalassistant used to manage SIP-based communications. IPA Agent General Personalassistant used for usual daily life activities. The personal assistant is able to learn the usual habits of its use Table 3.1. Available IPAs projects
• 11.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 11 assistant design. An intelligent assistant for athletes was proposed called iAPERAS. It aims to help nonprofessional athletes who cannot afford their own training team to obtain dedicated information about health indicators. An IPA that allows doctors to monitor elderly people is proposed. This IPA, called Health Pal, helps elderly people monitoring the health status autonomously, interacting with devices such as thermometers, blood pressure measurement systems, and PDA applications. The medical records are stored on the user’s PDA as well as in the physician’s PC. Still on the healthcare field, there can be referred personal agents that detect: 1) The sleep quality of their users, such as the BeWell application 2) The mental well- being of college students, as demonstrated by the Student Life application on 3) The physiological indexes of elderly people illustrated by the PPCare agent. Li and Chen presented a framework for an intelligent assistant system based on case-based reasoning (CBR) and MAS for manufacturing. The MAS framework allows the creation of a collaborative environment among intelligent agents, decentralizing the processing in the system. Thus, the subsystems throughout the manufacturing system can be seen as individual, intelligent, and collaborative agents. Despite their usability, these personal assistants are usually used for leisure times. Different personal assistants are presented, where each one is able to operate on a specific domain (sports, healthcare, manufacturing, etc.). Despite their good usability on the specific fields they were proposed, these personal assistants cannot give a universal response to all the challenges that arise to their owners during their daily lives. The work available presents a good overview on the manner that IPAs can predict the behavior of their users. Nevertheless, its deployment in a real scenario is not completely demonstrated. Analyzing all the work that was already performed to develop personal assistants, it can be mentioned that it does not exist yet an optimized IPA able to answer automatically all the requests of their users. The answers that IPAs propose to their users can be improved by using bigger databases with information from more data sources.
• 12.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 12 CHAPTER4 IPAS AND THE IOTS Currently, IPAs are already capable to support their users in order to perform several tasks. It is necessary to ascertain whether it is possible to improve their knowledge and autonomy by using new technological paradigms. One possible solution would be the insertion of IPAs in IoT scenarios. It would enable IPAs to analyze information from a greater variety of sources, increasing the learning database. With the creation of IPAs that are more autonomous and insightful, many tasks performed by users can be eased. For example, considering a smart home scenario, where an IPA can communicate with other objects in the environment: a person wakes up at 7:00 am and must be ready to go working. As the alarm clock rings, the curtains in the user’s bedroom slowly open to allow some light to enter the room. Three minutes after the alarm clock stops, the person goes to the bathroom to take a shower. After the person takes a shower, at the kitchen, the toaster and the coffee machine turn on automatically to prepare breakfast. All these actions are transparent to the user and the IPA prepares the user’s morning routine autonomously. Some issues arise from the idea of introducing IPAs in IoT environments, mainly related to technological paradigms. One of these issues is related to the creation of mechanisms to support communication between objects within a network. One possible solution is the creation of a standard IoT communication protocol. Until then, other solutions can be attempted, such as the creation of network gateways to map different protocols and networks. The WSN is composed of 6LoWPAN nodes that use short addresses to communicate. The gateway can map these 6LoWPAN addresses into full IPv6 addresses, establishing the connection between the WSN and the Inter- net. Silva discussed a gateway that connects 6LoWPAN networks to IPv4/IPv6 networks. The gateway was used to map 6LoWPAN addresses to IPv4/IPv6 addresses and vice versa, building a ubiquitous communication scenario. Schrickte described the construction of a prototype for a gateway that connects IEEE 802.15.4 nodes with Ethernet IPv6 networks. The performance of the gateway was evaluated through its capability to compress and decompress IPv6 packets. With more objects communicating in the same environment, more data are generated. Some of that data may be important to the application scope, being necessary its management and storage. The necessary storage space will be enormous, so new mechanisms must be applied. The use of cloud computing technologies can be very useful in this case,
• 13.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 13 decentralizing the processing involved in data storage and management. Possible challenges arise from the lack of solutions to manage all the data generated in the IoT network inside the cloud, and the lack of a service to catalog the data generated by all the heterogeneous IoT devices in different categories. Zhou presented a cloud-based IoT architecture, describing a smart home environment containing PCs, PDAs, indoor and outdoor sensors, microwave ovens, toasters, televisions, and stereos, among other appliances. The interaction between those objects leads to some automatic actions that interfere with the daily life of the people who live in the house. The cloud infrastructure is used to store data coming from the objects. Dores addressed the feasibility of the integration of an IoT network with the Internet using multiple cloud services. They also described various technologies that can enable the construction of IoT networks, such as WSNs, next-generation networks, and cloud computing platforms. Rao described how IoT and the cloud computing technologies can work together to deal with the huge amount of data that exists in the network. A solution where the sensing of a certain physical measure is considered, a service on the cloud was proposed. This allows the creation of a ubiquitous communication scenario, where the sensors are the interface between the physical world and the digital world. In a scenario, where an IPA can communicate with every object that surrounds it, extra care must be applied to prevent private information from falling into the wrong hands. The authors identified five IoT security issues related to 1) The correct identification of IoT objects 2) The wireless link insecurities originated by each object connected to the IoT network 3) The heterogeneity of network devices and their specific security needs 4) The confidentiality of the information flowing on the IoT network, and 5) The information processing security. All these issues can be applicable to scenarios where IPAs are inserted into ubiquitous communication environments, as IPAs can behave as smart objects in an IoT environment IPAs can be used in many fields for various purposes. For example, a solution is described that uses a wireless body sensor network to monitor the vital signs of multiple patients. The information is passed to the intelligent personal digital assistants (IPDAs) of the medical staff using the ZigBee/IEEE 802.15.14 standard. The IPDA must analyze the data to correctly determine the health status of the patients. Virone presented a WSN for health monitoring. The system includes wearable devices, with some of them placed on the patient body and others placed around the house. The medical staff is informed about the health
• 14.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 14 status of the patients through notifications received on their PCs and PDAs Table 3 shows how IPAs can be used on two important scenarios: healthcare and smart home. Healthcare and smart home scenarios are two of the main areas that can benefit from the integration of IPAs on IoT contexts. It can allow the creation of optimized tools to Feature Importance Network gateways Forward the traffic between the WSNs and the Internet. They can also be used to do the mapping between different protocols, such as IPv4 to IPv6 or vice-versa. Future IoT gateways may support other protocols, such as CoAP, MQTT, among others Cloud computing Can be used to store and manage the data generated on IoT networks. There can be created mechanisms inside the cloud to catalog into different categories the data generated by different type of sensors. Those mechanisms can ease the management and the accessibility to the available resource Security mechanisms IoT networks can present some security issues, such as the correct identification of each entity in the network, assurance of the data privacy, and inefficiencies on the protocol and network security. New cryptography algorithms, key management systems, and security protocols can be needed to optimize the security on IoT networks. Table 4.1. Possible solutions to solve IoT network issues Application Usability Healthcare IPAs can be used on ambient-assisted living scenarios, where sensors monitor a person that is under care of a caretaker (medical staff or not). The caretaker uses an IPA that receives information from the body sensors. It allows a person to be monitored anytime and anywhere. The sensors can communicate through low-energy protocols (IEEE 802.15.4 or ZigBee). Smart home IPAs can be used to control and manage smart home appliances through mobile devices, such as smart phones. They can be used to turn off the lights inside a house, for example. Besides that, an IPA must be able to notify its user if some unexpected event is detected, such as a fire in the kitchen, or an abnormality on the air conditioner temperature. Table 4.2. Usability of IPAs on healthcare and smart homes
• 15.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 15 monitor patients in real time by giving detailed information about health indicators. On smart home scenarios, IPAs can be used to manage all the home appliances inside a house, anytime and anywhere. The future can lead to the creation of IPA agents capable to operate under multiple scenarios with good performance, serving as a tool to solve many of daily life tasks.
• 16.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 16 CHAPTER 5 DISCUSSION AND OPEN ISSUES IoT will allow scenarios of ubiquitous communication by connecting every device in a common network. This means that devices that should interact directly with humans, such as IPAs, can also be included in IoT environments. By interacting with other objects, IPAs can gain better knowledge about their users and consequently offer better assistance to them.. The creation of a standard protocol to be used by all devices, including power-constrained ones would enable the creation of ubiquitous communication scenarios with millions of devices. There are various candidates, such as CoAP, XMPP, and MQTT. These protocols are suitable for use in large networks based on small and low-power nodes, but they differ in some important aspects. Power consumption is one of the key aspects of IoT objects. With this in mind, among CoAP, XMPP, and MQTT, the most suitable protocol for the IoT would be CoAP, as it is the only one that uses UDP. MQTT has the MQTT-S version that uses UDP, but its architecture is more complex, as it requires specific gateways and brokers. Furthermore, UDP is a connectionless protocol, making it more suitable for larger networks. Security is another factor that is very important for IoT networks (and others). In terms of security, TCP is better than UDP, as the Transport Layer Security (TLS) and the Secure Sockets Layer (SSL) are deployed on top of TCP. UDP also provides security mechanisms with the usage of the DTLS but it is heavier than TLS. XMPP and MQTT can provide better security mechanisms than CoAP for IoT networks. Additionally, CoAP and MQTT support different QoS levels, while XMPP does not provide any QoS option. It is clear that a definite answer regarding the choice of an IoT protocol cannot be easily obtained. Table IV shows some important characteristics of the CoAP, XMPP, and MQTT protocols. The introduction of IPAs in IoT networks will create countless possibilities for services that can improve human life, but some issues need to be overcome. Some of the issues are related to IoT networks, while others are related to IPAs. Some open issues are as follows: 1) Standardization of a communication protocol for IoT devices. 2) Creation of mechanisms to store all the data generated in IoT networks. 3) Improvement of the security mechanisms used in the network, considering the growing number of communicating objects, which increases the potential sources of cyber attacks. 4) Validation of devices that want to join the network. IoT networks should respect the condition of the low level of users’ involvement in devices configuration. Therefore, users
• 17.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 17 should perform very few configuration on their IPAs, meaning that IPAs should be self configurable. This fact highlights the necessity to create a transparent mechanism that allows users to validate and manage the smart objects in the network. 5) Creation of better speech recognition mechanisms to ad- dress all the current problems that IPAs present to recognize voice commands, such as the environmental noise. 6) Improvement of the context awareness on IPAs. With these issues correctly addressed, IPAs can be used on different contexts, giving humans the opportunity to have a smart personal assistant to complete almost every task that they need to perform. Protocol Messaging Transport QoS Support Security CoAP Request/Response UDP YES DTLS XMPP Request/Response Publish/Subscribe TCP NO TLS/SSL MQTT Publish/Subscribe TCP, UDP (MQTT-S) YES TLS/SSL Table 5.1. CoAP, XMPP, MQTT Characteristics
• 18.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 18 CHAPTER 6 CONCLUSIONS IoT technology will enable the creation of ubiquitous communication scenarios, where almost all the devices in the environment will be able to communicate. A standard protocol for enabling communication in IoT networks would ease the creation of more ubiquitous communication solutions. Nevertheless, the available protocols, such as CoAP, XMPP, MQTT, and other proprietary protocols, allow the development of interesting IoT projects, such as smart city scenarios and smart home appliances. IoT technology offers new opportunities, such as the creation of IPAs that can assist their users while communicating with other smart objects in the environment. It can increase the knowledge of IPAs, as they would learn the behavior of their users through direct interactions with them and through inter- actions with other smart objects in the environment. To achieve this goal, it is necessary to develop new mechanisms to in- crease the intelligence of IPAs, such as new machine learning algorithms and speech recognition mechanisms. The inclusion of new types of devices in the network, such as devices with embedded sensor units, will increase the heterogeneity of the network, increasing the generated traffic as well. To address these issues, it is necessary to take into account the creation of new security mechanisms to reduce cyber attacks and problems related to the entity authentication. Another issue is related to the creation of mechanisms to manage all the data generated in the network. The creation of a scenario where machines interact with humans in a smart and fully aware environment may not be as far away as it is thought.
• 19.
  Intelligent Personal Assistantsbased on Internet of Things Approaches Dept. of electronics and communication 19 REFERENCES 1. B. Pediredla, K. I.-K. Wang, Z. Salcic, and A. Ivoghlian, “A 6LoWPAN implementation for memory constrained and power efficient wireless sensor nodes,” in Proc. 39th Annu. Conf. IEEE Ind. Electron. Soc., Vienna, Austria, Nov. 10–13, 2013, pp. 4432–4437. 2. J.J.P.C.Rodrigues and P.A.C.S.Neves, “A survey on IP-based wireless sensor network solutions,” Int. J. Commun. Syst., vol. 23, no. 8, pp. 963– 981, Aug. 2010 3. C. Qiang, G. Quan, B. Yu, and L. Yang, “Research on security issues of the internet of things,” Int. J. Futur. Gener. Commun. Netw, vol. 6, no. 6, pp. 1–10, Dec. 2013.