Coupled Layer Architecture for Robotic Autonomy

As part of CLARAty (Coupled-Layer Architecture for Robotic Autonomy), this framework shares the design aims of maximizing code reuse while maintaining an efficient and accessible implementation.

CLARAty is designed to ease the transition from research to software of flight-ready. It attempts to achieve this aim by developing a set of standard interface and a basic set of reusable components. CLARAty is being developed using object oriented design principles to provide an avenue and to enable code reuse for extension. An open source development model is being used to allow collaborators to contribute components extension, which helps to maintain a critical mass and architecture achieve.

One novel feature of the CLARAty architecture is its two layer structure, the top layer called as decision layer provides a combination of operational executive and procedural planner. The lower level called as functional layer provides a hierarchical interface to hardware components and rover services. The decision layer may access services in the functional layer at any point in the hierarchy, allowing the decision layer to plan at a granularity appropriate for a given task.

The motivation for developing a generic navigation framework comes from the experiences navigation algorithms implementation for a variety of robots. For instance, a combination of a local obstacle avoidance algorithm and a real time path planner has been used on a number of robotic platforms. The first implementation was developed for Ratler. It has been used on progression of robots including Nomad, an ATRV, and most currently Hyperion.

Every new implementation has made gains in performance and capabilities but a major effort has been required to port the software, often involving a complete reimplementation. A target of this work is to simplify this process, allowing researchers to focus on developing and testing new capabilities rather than dealing with the mundane details of creating a specific platform implementation of an existing algorithm.