Acceleration of Publish/Subscribe Messaging in ROS-compliant FPGA Component.

Intelligent robots demand complex information processing such as SLAM (Simultaneous Localization and Mapping) and DNN (Deep Neural Network). FPGA (Field Programmable Gate Array) is expected to accelerate these applications with high energy efficiency. Introducing FPGA into robots is difficult due to its high development costs. In order to introduce FPGA easily to a system on ROS (Robot Operating System) development platform, ROS-compliant FPGA Component has been proposed. However, large communication latency between ROS components is a severe problem. This research aims to reduce the latency by implementing Publish/Subscribe messaging of ROS as hardware. Based on result of network packets analysis in ROS system, we propose a method of implementing a Hardware ROS-compliant FPGA Component. It is done by separating registration part (XMLRPC) and data communication part (TCPROS) of Publish/Subscribe messaging. Evaluation result shows that the proposed Hardware ROS-compliant FPGA component can reduce delay time of Publish/Subscribe messaging even with small amount of hardware.