Design And Implementation Of Software Platform Based On UPnP Middleware For Service Robot Intelligent Space

The combination of service robot and intelligent space, not only expands the service robot’s capacity for environment perceiving and decision-making, but also enriches the ability of the intelligent space in the field of task performing. Service robot intelligent space is a heterogeneous, dynamic, and complex computing system. By introducing middleware technology into this field, the underlying complexity can be masked, thus the system overall scalability can be improved.This subject originates from a National Natural Science Foundation funded project, named "Human behavior understanding in service robot intelligent space" and a Shandong University Innovation Fund Project (Foster interdisciplinary proiect), named "Research on key technologies of the practicality of service robot". In the context of intelligent space, intelligent nodes such as sensors are distributed according to the characteristics and functional requirements of home environment; corresponding information is obtained through wireless sensor network, so that the service robot’s burden and cost can be reduced. This thesis proposes the service robot intelligent space software-platform as the structural framework, which undertakes the task of interoperating and communicating between intelligent nodes and intelligent space. The main contributions of this thesis are shown as follows:Firstly, by analyzing the characteristics and functional requirements of service robot intelligent space in home environment, this thesis proposes the combination of service-oriented architecture (SOA) and multi-agent model as the application structure model, and summarizes the principles of designing a software platform for service robot intelligent space. By investigating and evaluating various robot middleware, this thesis adapts a design scheme of the software platform based on UPnP middleware. The systematic architecture of the software platform consists of device abstraction layer, message-scheduling layer, and resource management layer. The interfaces between different layers are implemented according to the detailed functional requirements of the three layers.Secondly, according to UPnP standard, the functions provided by intelligent devices are abstracted as "services", and the devices are implemented by the combination of services. The original devices are converted to UPnP devices by using pupnp SDK as the software development kit for UPnP protocol. An UPnP converting module is implemented in OpenWrt OS running on D-Link DIR-615C2router, managing the various sensors and several intelligent devices. The original service robot was implemented in Windows OS, meanwhile, an UPnP wrapper is provided to convert it to an UPnP robot in this thesis.Thirdly, based on asynchronous messages, the messages involved in device interoperation are classified into ordinary messages, event messages, and stream messages. The subscribe/publish mechanism guarantees the loose coupling among intelligent devices, thus improved the robustness of service robot intelligent space.Fourthly, the resource management layer is presented as a central management node, which consists mainly of device management sub-module, task-scheduling sub-module. The device management sub-module ensures that the central management node always have the latest device information, especially when devices joining in or leaving the intelligent space. The task-scheduling sub-module gives the optimal task execution order in a global perspective of the intelligent space.Experiments and corresponding analysis verify the stable performance, high dynamic adaptability, low degree of coupling between devices, and scalability of the service robot intelligent space software-platform designed in this thesis. All the performance improved above lay a solid foundation for further research and practicality of service robot intelligent space.