Research Challenges in the Internet of Mobile Things

 

The number of intelligent devices continues to grow exponentially, giving these smart things the ability to sense, interpret, control, actuate, communicate and negotiate – over the hyper-connected Internet of Things (IoT) space. There are several reports that predict that by the year 2020 there could be in excess of 20 billion devices connected to the internetⁱ; and therefore, for understanding the complexities of this massive scaling, many research efforts are underway to model and design IoT-based systems.¹

Most of the research questions, however, have focused on systems with fixed sensors/devices deployments that are tagged to a particular location without the scope for any form of mobility. A new development in this regard has been the emergence of mobile things that has given rise to mobile IoT or the 'Internet of Mobile Things' (IoMT).^2,3 Examples include robots, unmanned aerial vehicles (UAV), and even commonplace moving objects such as cars, buses or trains. Even humans with wearable devices (such as smart watches, wrist bands, smartphones, etc.) would form part of the mobile IoT. Such objects would also remain connected to the internet even while moving around, giving rise to several innovation possibilities. As with all new innovation paradigms, such possibilities come with their own research challenges, including but not limited to the following:

1. Mobility induced sensor network design. The mobile IoT paradigm invalidates many of the assumptions of traditional wireless sensor networks,² especially with regards to wireless technologies and protocols. In particular, mobile IoT devices would find it quite difficult to connect with each other and other components of the IoT network in the presence of mobility, intermittent connectivity and RF link variability. For example, internet-connected carsⁱⁱ would need to transmit and receive data from different gateways depending on their location, which would keep changing due to their (the cars) mobility. This calls for radical new sensor network paradigms for IoT, perhaps borrowing ideas from mobile ad-hoc networks (MANETs)⁴ or delay-tolerant networks.⁵

2. Robustness. In case of fixed deployed sensing and actuation platforms, it is common for devices to know their locations, have synchronized clocks, have knowledge of their neighbors for cooperating, configure a consistent set of parameters (such as consistent sleep and wakeup schedules, appropriate power levels of communication, and security keys). However, in case of IoMT, the topology of the system will be highly uncertain and may vary sporadically. In such a case, maintainability of a long-lived and dynamic system is problematic, and raises challenges in areas such as device discoverability, power usage optimization and communication protocols.

3. Coordination. IoMT brings together different kinds of mobile sensing opportunities: (i) opportunistic that piggybacks on the mobility of phones or vehicles, and (ii) controlled that provides a precise account of where to sample the environment (e.g., using UAVs). In the latter case, deployment and programmability of multiple UAVs for collaborative sensing tasks is non-trivial. The dominant practice is to create a set of pre-defined commands for each UAV. However, an increase in simultaneously-deployed UAVs renders this approach unmanageable as programmers must manually decompose the goal into a set of single-UAV parallel tasks. Moreover, they must estimate the duration of each task and balance the load between UAVs to fully exploit parallelism, while taking into account temporal constraints. The timing of a UAVs action depends on unpredictable factors (such as obstacles and crashes), which complicate this time analysis. Hence real-time coordination among mobile sensing and actuation platforms is a crucial research challenge that needs to be addressed if IoMT is to become successful.

4. Concurrency. In many application scenarios, communication among many types of mobile sensing entities needs to happen in real-time rather than in a delay tolerant manner. The run-time system needs to consider each such platform individually instead of treating all equally as they require individual, but coordinated planning. Therefore, dealing with parallelism and concurrency is markedly different. In mainstream sensor networks, it is network communication and input/output operations that need to occur in parallel with data processing. Hence while concurrency is an issue in IoT in general, it is particularly complex in IoMT due to the inherent unpredictability of the navigation time and movement patterns of mobile entities. For example, an internet-connected car moving through unpredictable traffic could exhibit highly variable mobility patterns and travel times between its source and destination; managing communication and input/output operations among multiple such cars – each of them with differing mobility patterns – would be challenging.

5. Modeling of mobility patterns. Designing a network comprising mobile sensing and actuation platforms has to take into account their mobility patterns. For example, motor vehicles travel in certain pre-defined patterns depending on the nature of the vehicle, e.g., commercial vehicles, public transport, private vehicles driven by commuters, etc. Modeling these patterns and understanding their variability over a period of time would yield valuable insights to solve the following research issues: what types of mobile sensors to install, which protocols would be most suitable, which types of sensing platforms can directly send messages to each other, which messages need to be routed through gateways, etc.

6. Sensor aggregation and virtual sensing. Due to large numbers of sensors in a network, it sometimes becomes necessary to aggregate multiple sensing streams as one 'virtual' sensor ⁶ that provides an integrated interface to other functionalities. An example of a 'virtual' mobile sensor could be aggregated traffic flow information on a highway that is derived from multiple vehicle sensors. This raises inter alia the following questions: which sensors can be aggregated (considering their mobility patterns); how is the aggregation to be implemented (given that mobile sensors – by definition – are mobile, and therefore, may not be always in proximity of each other); if a sensor moves out of range, can the aggregation be recovered through other means, i.e., is the aggregation mobility-tolerant; etc.

7. Optimal data capture and processing. A key issue in IoT systems is the enormity of data produced and transmitted on the network. Since most of this data may not be useful to any user, techniques for optimally filtering the data before storage and dissemination to users will emerge as a crucial research area. In our earlier work,⁷ we have explored the use of a goal-driven approach for data filtering. However, in case of mobile sensors (e.g., on internet-connected cars), the mobility patterns of the sensors that bring in the notion of time dependency would necessitate a rethink of our approach. Here too, the mobility of sensors would raise the same issue of mobility-tolerant aggregation as raised above. Hence what is appropriate data for users could depend on the time at which they request the data. Techniques such as spatio-temporal graphs⁸ may become relevant in such cases.

8. Location-based data storage and representation. Optimal storage and representation of IoT data is a crucial topic, given the data volumes that may need to be stored for viewing and future analysis. To that end, our earlier work⁹ proposed the idea of data storage and migration among multiple 'mini-Clouds'. Our approach, however, assumed a fixed configuration with predictable sensor network topologies and latencies. The introduction of mobile sensors invalidates that assumption, and calls for a rethink of the following: how to select the 'best' mini-Cloud to which a mobile sensor can send its data; how this selection can be made dynamic based on the mobile sensor's mobility pattern; how data sent from the same mobile sensor to multiple mini-Clouds can be represented in an integrated manner, and kept synchronized for the purpose of later retrieval and analysis; and how this representation can be extended to data sent from multiple related mobile sensors.

9. Service implementation via actuation. Users often need to make changes in the IoT network via actuation. Examples include: self-repair of resources, changing resource state, (re-) configuration, etc. Indeed, aggregating data from multiple sensors would also involve actuation. Typical actuations involving IoT resources are complex in nature, and need to be modeled as service compositions,¹⁰ with each service in the composition modeling an actuation command. In the mobile IoT, this complexity will be further increased due to the mobility of the IoT resources themselves. This would also raise issues of quality of service (QoS) in terms of meeting pre-defined SLAs on deadlines and other QoS parameters. A deeper investigation of integration with network service composition¹¹ is therefore indicated.

10. Integration with opportunistic computing. The field of opportunistic computing¹² is concerned with how users with portable electronic devices can interact with each other opportunistically, and implement tasks on the fly. Mobile IoT further enriches this with the possibility of users interacting with (fixed or mobile) sensors, thereby giving rise to opportunistic IoT.¹³ Some examples could be: human-centric sensing, data sharing, forming opportunistic alliances based on mutual 'friendship' in social networks, on-demand community formation (disaster response, flash mobs), etc. The key research challenge here would be to determine the most optimal approaches to facilitate decentralized opportunistic interactions among human users and the IoT network.

We believe that IoMT has the potential to be a major game changer that will drive innovation in both traditional and nontraditional application domains. Therefore, in this article, our attempt has been to identify its core technical problems where mobility adds another complex dimension to existing spatiotemporal systems. While the compiled list is not exhaustive, it outlines the top ten challenges that need to be considered to potentially leap-frog into more advanced IoT outcomes.

 

References

1. IoT European Research Cluster. http://goo.gl/hpvhQo .

2. Luis E. Talavera, Markus Endler, Igor Vasconcelos, Rafael Vasconcelos, M. Cunha, and Francisco Jos'e da Silva e Silva. The mobile hub concept: Enabling applications for the internet of mobile things. In 2015 IEEE International Conference on Pervasive Computing and Communication Workshops, PerCom Workshops 2015, St. Louis, MO, USA, March 23-27, 2015, pages 123–128, 2015.

3. Klara Nahrstedt. Internet of mobile things: Challenges and opportunities. In Proceedings of the 23rd International Conference on Parallel Architectures and Compilation, PACT '14, pages 1–2, New York, NY, USA, 2014. ACM.

4. Carlos de Morais Cordeiro and Dharma P Agrawal. Mobile ad hoc networking. Center for Distributed and Mobile Computing, ECECS, University of Cincinnati, pages 1–63, 2002.

5. Kevin Fall. A delay-tolerant network architecture for challenged internets. In Proceedings of the 2003 conference on Applications, technologies, architectures, and protocols for computer communications, pages 27–34. ACM, 2003.

6. Sanem Kabadayi, Adam Pridgen, and Christine Julien. Virtual sensors: Abstracting data from physical sensors. In Proceedings of the 2006 International Symposium on on World of Wireless, Mobile and Multimedia Networks, WOWMOM ’06, pages 587–592, Washington, DC, USA, 2006. IEEE Computer Society.

7. Nanjangud Narendra, Karthikeyan Ponnalagu, Aditya Ghose, and Srikanth Tamilselvam. Goal-driven context-aware data filtering in iot-based systems. In Intelligent Transportation Systems (ITSC), 2015 IEEE 18th International Conference on, pages 2172–2179. IEEE, 2015.

8. Betsy George, James M. Kang, and Shashi Shekhar. Spatio-temporal sensor graphs (stsg): A data model for the discovery of spatio-temporal patterns. Intell. Data Anal., 13(3):457–475, August 2009.

9. Nanjangud Narendra, Koundinya Koorapati, and Vijayalaxmi Ujja. Towards cloud-based decentralized storage for internet of things data. In Cloud Computing for Emerging Markets (CCEM), 2015 IEEE International Conference on. IEEE, 2015.

10. Li Liu, Xinrui Liu, and Xinyu Li. Cloud-based service composition architecture for internet of things. In Yongheng Wang and Xiaoming Zhang, editors, Internet of Things, volume 312 of Communications in Computer and Information Science, pages 559–564. Springer Berlin Heidelberg, 2012.

11. Jun Huang, Guoquan Liu, Qiang Duan, and Yuhong Yan. QoS-aware service composition for converged network-cloud service provisioning. In Services Computing (SCC), 2014 IEEE International Conference on, pages 67–74, June 2014.

12. M. Conti, S. Giordano, M. May, and A. Passarella. From opportunistic networks to opportunistic computing. Communications Magazine, IEEE, 48(9):126–139, Sept 2010.

13. Bin Guo, Zhiwen Yu, Xingshe Zhou, and Daqing Zhang. Opportunistic iot: Exploring the social side of the internet of things. In Computer Supported Cooperative Work in Design (CSCWD), 2012 IEEE 16th International Conference on, pages 925–929, May 2012.

 

ⁱ http://www.informationweek.com/mobile/mobile-devices/gartner-21-billion-iot-devices-to-invade-by-2020/d/d-id/1323081
ⁱⁱ http://tf.nist.gov/seminars/WSTS/PDFs/1-0_Cisco_FBonomi_ConnectedVehicles.pdf

 

---

 

Dr. Nanjangud C Narendra (NCN) has joined Ericsson Research Bangalore as a Principal Engineer. Prior to joining Ericsson, he worked in MS Ramaiah University, Cognizant, IBM Research, HP India and Motorola India. He has about 22 years R&D experience in the Indian IT industry. His research interests span Software Engineering, Workflow Management, Web Services, Service Oriented Architecture, Cloud Computing and Internet of Things. He has published over 100 papers in international conferences and journals. He has been a Program Committee member for several key conferences such as Autonomous Agents and Multi-Agent Systems (AAMAS), International Conference on Service Oriented Computing (ICSOC), IEEE International Cloud Computing Conference; he was also on the Program Committee for ICSOC (Industry track) in 2012 and ICSOC (Research track) in 2015. He is a member of the Editorial Board of Service Oriented Computing and Applications journal. He is a Senior Member of IEEE and ACM.

 

Prasant Misra is a Scientist at TCS Innovation Labs, Bangalore. He performs research in low-power and energy efficient sensing/signal processing/wireless communication with a focus on system design within the general framework of Cyber Physical Systems and Internet of Things. He received his Ph.D. in Computer Science and Engineering from the University of New South Wales, Sydney. He has received many awards/honors in recognition of his work, of which it is noteworthy to mention the AusAID (Australian Government) Australian Leadership Awards (2008) and the ERCIM Alain Bensoussan / Marie Curie Fellowship (2012). He is a senior member of IEEE, member of ACM, secretary of IEEE Computer Society (Bangalore Section), and an Associate Technical Editor of IEEE Communication Magazine.