1.5) Experiment Environment Types and Execution

AERPAW Environments

The AERPAW platform has three experimentation modes: 

• Development (Digital Twin): The development environment involves virtual experiments over AERPAW's cloud resources. Experimenters have live access to this environment for their experiments and can develop their experiments here. In this environment, each ARN maps to an AERPAW Virtual Node (AVN).  Experimenters develop code on each AVN that they desire should run on the corresponding ARN.  A custom-designed emulation of radio environment and vehicle mobility is integrated into this environment, to enable the software being developed by the Experimenter to run with no disruption.  However, at this time, the emulation is not physically faithful to any particular radio or aerial environment; no meaningful data can be usefully extracted about any physical aspect of AERPAW's radio or aerial environments from executing in this environment.

Internally, we distinguish between a variant of this environment called the Emulation environment, which is identical to the above, but does not allow live access by the Experimenter; it is an instance of a Development environment that may be optionally utilized by AERPAW Operations personnel to validate or safety-check an experiment in a virtual environment.

• Sandbox: The sandbox environment involves real radio hardware in an indoor environment.  While the AERPAW Radio Nodes in this environment are real hardware, consisting of real USRPs and UAVs/UGVs, and are indistinguishable from field-installed AERPAW Radio Nodes, they are not installed in the field, but rather in labs with controlled surroundings.  The radio signal transmissions from the USRPs are limited, and the UAVs/UGVs are restricted from actual motion.  Experimenters have live access to the Sandbox environment for their own experiments, but this access will need to be scheduled. AERPAW Sandbox will be made generally available at the end of Phase-3; until then the Sandbox mode is unavailable to Experimenters.

• Testbed: In this mode, the experiments are executed over actual AERPAW fixed and portable radio nodes in the field. AERPAW is a batch-processing facility; Experimenters do not have live login access to the Testbed environment (except to Keysight RF sensors).  The experiment code developed by the Experimenter in the Development mode is executed by AERPAW Operations personnel in the Testbed mode, after the Experimenter submits the experiment for Testbed execution.

Experiment Execution

Experiments execute in one of several environments as described above.  Each experiment is considered a distributed program executing on a specific set of AERPAW Radio Nodes (ARNs) in the Testbed environment, as selected by the Experimenters.  Each ARN represents some (i) local (edge) compute resources, (ii) radio resources, such as Software Defined Radios (SDRs), and (iii) optionally, vehicle resources, such as a programmable Unmanned Aerial/Ground Vehicle (UAV/UGV).   ARNs are briefly described further below, and in more detail later in this User Manual.  Experimenters have the opportunity to code this program, on a node-by-node basis, including SDR control code, vehicle control code, test traffic generation and capture code, and any other arbitrary experiment logic the Experimenter might wish to code.

In AERPAW,  ARNs are the only locations where Experimenter code runs.  An Experimenter planning to represent some real-world scenario of interest to an AERPAW Experiment must first map all the compute locations of interest in that scenario to one or more ARNs in AERPAW.  For example, to represent a real-world scenario with one cellular base station and one mobile User Equipment (UE) device, one must design an AERPAW Experiment with one ARN to run a software implementation of a base station, and another ARN with software implementation of a UE.

When an Experiment is instantiated in the Development Environment, the AERPAW Backplane creates two distinct dedicated Layer-2 networks for it, isolated from other Experiments, and AERPAW facility management.

• The first connects all the AVNs in the experiment - this is the Experimenter's Management (XM) network, that is intended to provide the Experimenter-authored code to coordinate operation of Experiment logic across AVNs.  This is often part of experiment design, to allow the Experimenter to predictably experiment with scenarios that arise rarely or by chance in the real-world scenario.  For example, the Experimenter may be interested in studying the case where the base station suffers a temporary high processing delay just as a UE controlling a UAV has moved to a certain altitude.  To ensure repeatable occurrence of this exact scenario, which in the real-world arises by chance and does not involve any communication between the UE and the base station, the Experimenter would use a coordination message sent by their experiment logic on the UE when the target altitude is reached, using the XM plane. The XM plane also allows live login access to each AVN during Development.  This ability does not persist in Testbed Mode, but the XM plane itself is re-created on the ARNs, so any such coordination messages generated by the Experiment logic in various AVNs would be successfully exchanged.

• The other connects only those AVNs of the Experiment that represent fixed (not portable) nodes.  Such nodes, in the target real-world scenario, may well be connected through a wired network, to allow them to exchange coordination messages that are part of the real-world scenario.  For example, in an Experiment that includes multiple ARNs representing multiple cellular base stations, these base stations are connected by a core network.  In the AERPAW Experiment representing this scenario, the corresponding AVNs would be connected by an Experimenter's Data (XD) network; this network would persist into the Testbed environment, connecting the corresponding subset of ARNs.  At this time, each fixed node is provided with a single XD interface, which are all connected in a flat Layer-2 network.