The TRADE Middleware for Advanced Robotic Architectures

Abstract

Over the last decade, the Robot Operating System (ROS) has become the de facto standard for robotic middleware having significantly improved some of the shortcomings of version 1 with the release of version 2. Yet, while the focus of ROS 2 has been “downward” on the underlying communication layer, the interfaces “upward” to the robotic architectures implemented in ROS has received little attention. In this paper, we argue that robotic middleware can serve important roles for robotic architectures, in particular, cognitive robotic architecture, if the right kinds of interfaces are provided that allow for a tight integration between architecture and middleware. We introduce the Thinking Robots Agent Development Environment, TRADE, which is an extension of the previous Agent Development Environment, ADE, and provides advanced features for architecture integration and interactions between cognitive robotic architecture and the middleware layer. We describe several features in TRADE that are missing in ROS, in particular, system-wide locking mechanisms, service instrumentation, and middleware service calls and discuss how they can support the architecture developers with implementing advanced architectural features such as dialogue-based system debugging and configuration, or multi-effector multi-robot behavior coordination.