1 Spinning Sensors: Middleware for Robotic Sensor Network Soko Aoki * , Yukihiko Kirihara † , Jin Nakazawa * and Hideyuki Tokuda * * Graduate School of Media and Governance, Keio University, Kanagawa, Japan Email: {soko, jin, hxt}@ht.sfc.keio.ac.jp † Triple Double Corporation, Tokyo, Japan Email: kirihara@triple-double.co.jp Abstract—This paper proposes Spinning Sensors, a middleware system for creating robotic sensor network applications. This middleware enables application programmers to easily write application software which utilizes both sensors and actuators in their network. In the Spinning Sensors model, a sensor node is attached to a robotic actuator to change its position and/or di- rection. The continuous position change enables the single robotic sensor node to cover greater sensing area and sensing time. This paper describes robotic sensor node model that the sensor and the actuator are attached together as one node, then present our design and implementation of the software which achieved both versatility and functionality simultaneously. We constructed three applications using the middleware: an environment monitoring, a radio control robot, and context-aware services. We show the result of experiment in which we utilized multiple robotic sensor nodes to monitor the environment. KEYWORDS Sensors, Actuators, Robotic Sensors, Sensor Network, Co- ordination, Middleware, Robotic Interactions with Sensors I. I NTRODUCTION A variety of sensor-based ubiquitous computing applica- tions, such as environment monitoring, industrial monitoring, and context acquisition are proposed. The major issue in these applications is achieving maximum effectiveness (e.g. coverage and granularity of sensing targets) with minimum consumption (e.g. the number of sensors and maintenance cost). One approach to construct a sensor network system with large coverage and high granularity is to deploy as many sensor nodes as possible in the environment. This is feasible in case the sensor node is cheap. The other approach is to make a sensor node mobile so as to change its coverage area dynamically. This approach is feasible when the sensor node is expensive or when the administrator wants to reduce the number of the sensor nodes. We propose Spinning Sensors system that increases cover- age of a sensor node, and decreases the number of sensors required in an application by realizing a robotic sensor node. In the system, a sensor node is attached to a robotic actuator in order to move or rotate the sensor node to follow moving objects or to increase its coverage, respectively. The three major features of Spinning Sensors are hardware abstraction of sensors and actuators, communication and coordination mech- anism for sensors and actuators, and application programming interface for developers. As the newest information system is composed of het- erogeneous general purpose information devices, the sensor network system will be composed of general purpose sensors and actuators in the near future. Since there are many kinds of sensors, actuators, and robots in our network, the Spinning Sensors needs to hold versatility and functionality simultane- ously. In terms of versatility, we have divided the hardware control software into two classes: abstract class and imple- mentation class. With this design, we could separate hardware specific programming and general purpose middleware. The communication and coordination mechanism is provided by this middleware layer so that the application programmers can easily construct multiple robotic sensor nodes environment. The contributions of this paper are the following. Spinning Sensors showed the concept of putting together general pur- pose sensors and actuators to increase the sensing coverage and sensing granularity by showing the design and implementation of Spinning Sensors. The middleware of Spinning Sensors realized both versatility and functionality by adopting modular design of the software so that the system can be utilized for many kinds of hardware devices and applications. Finally we could show the result of implementation of Spinning Sensors by showing the experiment data telling that there is an improvement in the quality of sensed data by using the Spinning Sensors model. The rest of the paper is organized as follows. Sections 2 categorizes the conventional sensors and actuators and explain the necessity of the robotic sensor node which combines general purpose sensors and actuators as one node. Section 3 presents the design and implementation of Spinning Sensors. In section 4, we show the result of experiment using multiple robotic sensor nodes. Section 5 surveys related work and section 6 concludes this paper. II. ROBOTIC SENSOR NETWORK In this section, we classify the sensors and actuators by their characteristics. Sensors can be classified by two criteria: sensing range and sensing direction. Some of sensors cover a few meters (long range) and some others cover only a few centimeters (short range). The sensing angles also vary. For example, thermometers and hygrometers cover 360 degrees (non-directional) in a sense that they measure the condition of surrounding air. Contrarily, illuminometers, cameras, and