Ville Könönen

Automatic feature selection for context recognition in mobile devices

In mobile devices there exist several in-built sensor units and sources which provide data for context reasoning. More context sources can be attached via wireless network connections. Usually, the mobile devices and the context sources are battery powered and their computational and space resources are limited. This sets special requirements for the context recognition algorithms. In this paper, several classification and automatic feature selection algorithms are compared in the context recognition domain. The main goal of this study is to investigate how much advantage can be achieved by using sophisticated and complex classification methods compared with a simple method that can easily be implemented in mobile devices. The main result is that even a simple linear classification algorithm can achieve a reasonably good accuracy if the features calculated from raw data are selected in a suitable way. Usually context recognition algorithms are fitted to a particular problem instance in an off-line manner and modifying methods for on-line learning is difficult or impossible. An on-line version of the Minimum-distance classifier is presented in this paper and it is justified that it leads to considerably higher classification accuracies compared with the static off-line version of the algorithm. Moreover, we report superior performance for the Minimum-distance classifier compared to other classifiers from the view point of computational load and power consumption of a smart phone.

Optimizing power consumption of always-on applications based on timer alignment

In modern smartphones there exist several applications that periodically establish data connection and/or send data to the network. For example, a mobile email application periodically checks for incoming messages from the server. Packet transmission is usually triggered by expiration of a timer. Wireless radio access resources are activated during transmission, which increases energy consumption of the terminal, particularly when connected over 3G network. Fortunately, timing requirements are usually not strict and therefore application timers can be grouped together by using a timer alignment method. In this paper, we present two timer alignment methods for reduction of energy consumption. In the first both minimum and maximum thresholds of timers are available for alignment, and in the second only the maximum timer value is available. One of the methods was tested in the real environment by using Nokia N97 mobile phone in two different cellular networks. In addition, both methods were tested with a large number of randomly generated timers to find out certain statistical properties of their behavior. The tests presented very promising results, maximal power savings were ∼50%. However, the results depend on certain parameters of the cellular network.

Decision Making Strategies for Mobile Collaborative Context Recognition

This paper focuses to a problem domain where several imperfect context recognition devices with restricted resources are located in the same context. The main aim of the paper is to compare different collaborative decision making methods in this domain. In particular, the focus is on certain voting mechanisms and confidence evaluation metrics. Also computational, space, and communication requirements of the discussed methods are discussed briefly. The methods are tested with a real-world dataset containing data from several sport activities. In addition, an online context recognition platform that is capable to make distributed context recognition decisions is introduced briefly.

Simulation and Implementation of the Autonomic Service Mobility Framework

The increased traffic load, proliferation of network nodes and, in particular, wireless user devices, and the boom in user services and exponential growth of information stored in content distribution networks (CDNs) have brought new challenges for current networks. One major challenge has and continues to be efficient load balancing and information access. The topics have been well studied for wired networks for, for example, process load balancing in distributed computer networks with migration. However, the wireless networks and ubiquitous computing environments create new limitations and additional requirements to perform service or process migration. In this paper, we present a simulation case and proof-of-concept implementation for service mobility as a part of the BIONETS service evolution process with the aim of optimizing service penetration in a pervasive computing environment and balancing the load in the system caused by the high service utilization rate.

Anatomy of automatic mobile carbon footprint calculator

A number of web-based systems provide environmental information to support our ecological lifestyle. However, there exists a need for more accurate and automated online systems facilitating personalised ecological awareness. This paper presents a travel type detection system for an automatic mobile carbon emission calculator. The system is able to detect automatically trips made by mobile users and provide them an easy and quick way for estimating the travel related CO2 emissions. The system is based on energy and computationally effective multisource fusion in a mobile device. The paper introduces the system with design rationale, explains smartphone implementation and provides comprehensive evaluation from both accuracy and energy efficiency points of view. The system attained the overall travel type detection accuracy of 77% and the daily average current consumption of 33mA. The results show great promise for the developed methodology to facilitate 24/7 carbon emission estimation of a traveller.

Optimizing meeting scheduling in collaborative mobile systems through distributed voting

Decision making through distributed voting can help automate routine-like collaborative tasks. In this paper we concentrate on how distributed voting strategies can be used for scheduling meetings in mobile and pervasive environments. Our work focuses on optimizing the meeting scheduling result for each participant in a mobile team by using user-specific preferences and information available on their devices. This negotiation is done in a distributed manner directly between the peers. In this paper we describe different approaches for the decision making strategy involving voting theory to balance out the different user preferences and availabilities. The weight of the votes from each participant can also be adjusted according to their importance or necessity in the given meeting. We also introduce briefly an approach to support distributed decision making strategies pervasively using a lightweight Web-based platform. To conclude the paper, we give our views on the future development directions and evaluation plans as well as extend the approach for other related domains.

Collaborative Context Recognition for Mobile Devices

The next wave of mobile applications is at hand. Mobile phones, PDAs, cameras, music players, and gaming gadgets are creating a connected mobile ecosystem where it is possible to implement systems with significant embedded intelligence. Such advances will make it possible to move many functions of the current PC-centric applications to the mobile domain. Since the inherent difficulties that come with mobility—limited UIs, short attention spans, power dependency, intermittent connectivity, to name but a few—are still not going away, new solutions are needed to make mobile computing satisfactory. We are facing the paradox of cramming ever more functions into our ever more portable devices, while seeking to achieve radically better usablility and semi-usable automated intelligence.

Architecture Requirements of the Future Internet.

The current Internet architecture curls around an original conversational model developed in the 1970’s. New solutions were for a long-time built on that framework. However, the current architecture is not able to meet optimally challenges posed by new access technologies, applications and services any more. In this paper, the purpose is to bring a new clarity to the architectural issues of the Future Internet by abstract reasoning. The paper explains the importance of architecture and introduce an architecture gap (problems and bottlenecks) in the current Internet as well as some essential requirements for a new architecture. The choice of requirements set by emerging and new challenges may be the most critical issue determining the new architecture. Therefore, the development of the new architecture should be guided by an understanding of the requirements. On the basis of requirement analysis we also propose a new architecture framework, which aim is to provide greater functionality, lower costs and increased adaptability for different types of communication for the Future Internet.

A computationally light classification method for mobile wellness platforms

The core of activity recognition in mobile wellness devices is a classification engine which maps observations from sensors to estimated classes. There exists a vast number of different classification algorithms that can be used for this purpose in the machine learning literature. Unfortunately, the computational and space requirements of these methods are often too high for the current mobile devices. In this paper we study a simple linear classifier and find, automatically with SFS and SFFS feature selection methods, a suitable set of features to be used with the classification method. The results show that the simple classifier performs comparable to more complex nonlinear k-Nearest Neighbor Classifier. This depicts great potential in implementing the classifier in small mobile wellness devices.

A Learning Automata Approach to Multi-agent Policy Gradient Learning

The policy gradient method is a popular technique for implementing reinforcement learning in an agent system. One of the reasons is that a policy gradient learner has a simple design and strong theoretical properties in single-agent domains. Previously, Williams showed that the REINFORCE algorithm is a special case of policy gradient learning. He also showed that a learning automaton could be seen as a special case of the REINFORCE algorithm. Learning automata theory guarantees that a group of automata will converge to a stable equilibrium in team games. In this paper we will show a theoretical connection between learning automata and policy gradient methods to transfer this theoretical result to multi-agent policy gradient learning. An appropriate exploration technique is crucial for the convergence of a multi-agent system. Since learning automata are guaranteed to converge, they posses such an exploration. We identify the identical mapping of a learning automaton onto the Boltzmann exploration strategy with an suitable temperature setting. The novel idea is that the temperature of the Boltzmann function is not dependent on time but on the action probabilities of the agents.