Data Flow¶

Rather than look at all data flowing through the system at once, it's easier to look at diagrams for specific data flows. On this page we provide diagrams for the 2 most critical data types, historically: navigation points and pointclouds.

Navigation Points¶

We use 3 different types of navigation points: Goals, targets, and setpoints. Each indicates increasing granularity. For example, a goal could be a pin dropped on the frontend map (called SETPOINT mission in the interface, maybe we should rename it). Then short-range targets are created in path manager as subgoals. This is because path planning and obstacle avoidance is more effective on short distances. This subgoal could be sent directly to path planner, but if explorer is enabled, first path manager will try to find a destination near the subgoal but far from obstacles (or whatever other decision-making criteria you implement, see Decision-Making). Either way, the subgoal is the target for the path planner. When path manager receives the resulting path, it breaks that down into even smaller MAVROS-ingestible setpoints, called such because the MAVROS topic is /mavros/setpoint_raw/local.

flowchart TD
    subgraph Frontend
        A[User Input / Mission Upload]
    end

    subgraph "REST Server"
        B[Convert to ROS JSON string]
    end

    subgraph "Task Manager"
        C[State machine
        Mission prep]
    end

    subgraph "Path Manager"
        E[Breakdown subgoals
        Terrain adjust
        Path into MAVROS setpoint]


        subgraph Explorer
            D[Find open space
            Select best nearby target]
        end
    end

    subgraph MAVROS
        F[Setpoint execution]
    end

    subgraph "Path Planner"
        G[Find path]
    end

    A -->|mission| B
    B -->|mission| C
    C -->|goal| E
    E -->|goal| D
    D -->|clear space target| E
    G -->|path to target| E
    E -->|target| G
    E -->|setpoint| F

For example, the goal set for the mission could be:

Then the path manager breaks this into subgoals which are sent to the path planner. In this image, explorer manager was enabled, so it produced 10 possible destinations (unselected options in green). The selected target is the pink ball. This was sent to the path planner, resulting in the green path. The path is then converted into MAVROS setpoints, as indicated by the red arrow just next to the drone (indicated by the TF stack) here.

Pointclouds¶

The pointcloud goes through a number of transformations, and almost every downstream perception node should be using the final pointcloud from the dataflow, called /cloud_aggregated. This is the cloud used by path planning, powerline detection, and trail following.

flowchart TD
    subgraph Drone
        Livox[Livox]
        Jetson[Jetson Orin Nano]
    end

    subgraph Jetson
        Livox_wrapper[Livox ROS Wrapper]
        PCL[PCL Analysis]
        SLAM[Fast LIO SLAM]
        TaskMgr[Task Manager]
        Perception[Multiple ROS2 perception nodes]
    end

    Livox -->|raw data|Jetson
    Livox_wrapper -->|raw livox custom pointcloud|SLAM
    SLAM -->|livox frame registered Ros2 Pointcloud2|TaskMgr
    TaskMgr -->|map frame registered Ros2 Pointcloud2|PCL
    PCL -->|aggregated and cleaned Ros2 Pointcloud2|Perception