Edge/Fog/Cloud Model for Computing in Robotics

Edge Computing

Processing and Storage of data closest to the source of data in the robot (sensors and controllers of the robot) are called edge computing. This typically means the computational capability the robot has onboard itself. The proximity of the edge to the source means the decision making on this is instant, but since the edge compute capability must be placed on the robot itself they are size, power, storage and thermally limited. To compensate they have good connectivity in the form of WiFi (indoor applications) or 4G or 5G modems (Outdoor applications) for connection to other computation. Examples of edge compute devices are NVidia Jetson series, Intel NUC series etc. These platforms offer good connectivity to servers, good power to performance ratio to be able to be used as onboard compute platforms.

Computing

It can be viewed as the intermediate layer between edge and the cloud. It is usually imagined as connected to the edge devices via LAN(Local Area Network). This can be viewed in the form of a base station for a robot: A fixed computer which the robot communicates to present on the LAN and mediates communication to the more powerful cloud servers. They are typically used for coordinating multiple edge devices like multiple robots, the fog computer allows for better robot-robot co-operation. This stage can be as powerful as the system designer wants it to be. It has some latency as the data from the robot has to be received by wireless means but better latency than on cloud.

Cloud Computing

This refers to large servers not present in the vicinity of the robot but the robot communicates via the internet. These are typically the most high power and high performance computers. These can be used for large amounts of computation and for logging and storing all the data the robot accumulates. Since these are central this can be used as a communication hub for multiple robot systems in different locations. Since the communication happens over the internet the latency is higher than the edge and fog cases. Typically large companies offer cloud computing with their servers for example Amazon Web Services, Microsoft Azure, Linode are some examples of providers of cloud computing. We can rent some servers of required processing and storage capacity for our application.

Leveraging this model

To leverage this model to make maximum use for us, we need to break down the tasks the robot needs to perform into how fast the response needs to be and also how processing heavy the tasks are. The tasks which require faster responses are best kept on the edge and fog whereas the processing heavy tasks are best allocated to the cloud.