Architecture

CARP is split into four parts:

• a platform-independent domain model (carp_core);
• a client framework that runs studies on the phone (CARP Mobile Sensing, CAMS);
• pluggable sampling packages;
• backends that deploy studies and store data (CARP Web Services, CAWS).

Components

flowchart LR
    subgraph Phone["Flutter app"]
        CAMS["carp_mobile_sensing<br/>(CAMS runtime)"]
        Pkgs["Sampling packages<br/>context, survey, health, audio,<br/>Apple Watch, Polar, Movesense, …"]
        Core["carp_core<br/>domain model"]
        Pkgs --> CAMS
        CAMS --> Core
    end
    CAMS -- "carp_backend<br/>carp_webservices" --> CAWS[("CARP Web Services<br/>(CAWS)")]
    Portal["CARP Portal"] --> CAWS

• Sampling packages, including those for wearables, plug into CAMS.
• CAMS builds on the carp_core domain model.
• CAMS communicates with CAWS through the carp_backend and carp_webservices packages.
• The CARP Portal connects to CAWS.

Study lifecycle

sequenceDiagram
    participant R as Researcher
    participant S as CAWS
    participant C as SmartPhoneClientManager
    participant P as Probes
    R->>S: Upload study protocol
    R->>S: Invite participant (study deployment)
    C->>S: tryDeployment()
    S-->>C: Primary device deployment
    C->>P: start() sampling
    P-->>C: Measurements
    C->>S: Upload data (data manager)

1. The researcher uploads a study protocol to CAWS.
2. The researcher invites a participant, which creates a study deployment.
3. On the phone, SmartPhoneClientManager calls tryDeployment(), and CAWS returns the primary device deployment.
4. The client calls start() on its probes, which return measurements.
5. A data manager uploads the data to CAWS.

Next steps