With embedded sensors –whether in urban infrastructures or in user’s mobile devices– being an integral part of future smart cities, it is natural to endeavor to acquire the information sensed by these devices to generate a view of the city’s current state. However, we argue that it is not sufficient to simply be aware of the physical state of the city. We need as importantly to be aware of its social state, which a) is difficult to sense compared to physical phenomena, b) critically involves participation by the citizens themselves in the sensing process, and c) requires privacy enforcement and data anonymisation techniques. A classic example would be the detection of the levels of crowd in public transit, and the resulting discomfort to passengers. While the former can be potentially sensed through combining data from turnstiles, motion sensors, and cameras, it is much cheaper to combine data from the citizens currently in transit, equipped with mobile connected devices. However, the latter is more difficult to sense passively, and requires active participation of the users.
Building upon the research of the MIMOVE team, the objective of this effort is to address the issues involved in creating a software infrastructure that can be used to adequately sense the physical and social phenomena in the city through user devices, both passively using embedded sensors, as well as through active citizen participation. The research performed in this area will be incorporated into a middleware using which urban physical and social sensing applications can be developed. In addition, the SMIS team contributes to the elicitation of the middleware solutions with research on privacy and security by design.
State of the art and challenges
We envision the following challenges:
- Heterogeneity: Enabling urban scale physical and social sensing and actuation will involve an unprecedented challenge in terms of the variety of a) embedded devices and platforms hosting the physical sensors and actuators, b) service interfaces and data formats of available information, and c) online social networking and other communication platforms preferred by the citizens. While the problem of interoperability among sensing services has been tackled in general in the literature, the specific challenges arising due to the mix of physical and social sensing, as well as the increased proliferation of both smart phone operating systems and social networks remain to be addressed.
- Scale, dynamicity and privacy: Future cities will easily contain tens of millions of active connected sensing devices, generating data actively as well as through the participation of the tens of millions of citizens using them to report on the happenings in the city. Most approaches to this problem however delegate the scale issue to cloud-hosted infrastructures, with only a few recent efforts to address the scalability arising at the device level itself [Hachem13]. While the current approach works, device-level scalability approaches will help avoid hitting a telecom-operator-capacity wall that we will arrive at in the future. In addition, we will follow a privacy by design approach to enforce privacy. As a related issue, it is unreasonable to assume that all these devices (and people) will be either available or fixed. Handling the resulting dynamicity thus becomes an integral challenge towards realizing such a platform.
- Fixed vs mobile sensing: Mobile sensing has significant advantages over fixed sensor infrastructures and especially the ease of deployment. However, the mobility of sensor nodes require revisiting the sensing systems that have been largely devised for fixed sensor network so that similar functionalities may indeed be offered.
- User motivation: Since users will play a big role in these systems, and their resources, including (mobile device) battery power and personal attention, are at a premium, adequate incentive mechanisms need to be developed to enable the collection of physically and socially sensed data. This has been the subject of interest in the participatory sensing domain [James12], but no silver bullet has so far been found. As mobile devices and social networks move from being used by a few to by everyone in future cities, powerful, potentially customized incentive mechanisms need to be developed.
Our work to address the above challenges will be in the following steps:
- Models for physical and social data, context, and interactions: Based on semantic techniques, we will propose extensible models of physical and social data and their sources, including those for various social networks and their access control mechanisms.
- Algorithms for scalable, mobile and adaptive sensing: We will design algorithms to play the dynamicity and scale challenges against each other, so that devices and people will be able to substitute those that get disconnected or otherwise become unavailable. For this, we will make use of mobility models as well as estimation techniques.
- Novel incentive models: To power the above we will explore techniques to encourage user participation, based not only on the users’ physical context (battery power, current location, time of day), but also their social context (with friends, in playful mood, etc.).
The above will be wrapped into a middleware platform, enabling developers to both collect physical and social data, as well as create innovative applications from the open availability of this data.
[Hachem13] S. Hachem, A. Pathak, V. Issarny. Service-Oriented Middleware for Large-Scale Mobile Participatory Sensing. In Pervasive and Mobile Computing Journal. 2014.
[James12] L.G. Jaimes, I. Vergara-Laurens, M.A. Labrador. A location-based incentive mechanism for participatory sensing systems with budget constraints. Pervasive Computing and Communications (PerCom). 2012.