Ringil 🗡️

GitHub

"Yet with his last and desperate stroke Fingolfin hewed the foot with Ringil ..."

Ringil is a system of autonomous swarms for objects--drones, submarines, etc. Decisions are events recorded and then determined by Galadril, but each entity can independently decide to take action to avoid an obstacle.

Caution

This project is still in its early stages.

Targeted architecture

flowchart TB
    classDef edge fill:#e1f5fe,stroke:#01579b,stroke-width:2px
    classDef onboard fill:#fff3e0,stroke:#e65100,stroke-width:2px
    classDef central fill:#f3e5f5,stroke:#4a148c,stroke-width:2px,stroke-dasharray: 5 5
    classDef action fill:#ffebee,stroke:#b71c1c,stroke-width:2px

    subgraph Galadril_Environment ["Central Intelligence (The Mirror)"]
        direction TB
        G_ESKG[("Galadril ESKG Cluster")]:::central
        G_Logic["Strategic Foresight / Policy"]:::central
    end

    subgraph Ringil_Entity ["Ringil Autonomous Node"]
        direction TB
        
        subgraph Perception_Layer ["Sensory Input"]
            S_Vision["Computer Vision / LiDAR"]:::edge
            S_Telemetry["Inertial / GPS"]:::edge
        end

        subgraph Edge_Brain ["Local Autonomy (The Instinct)"]
            Local_Avoidance{"Obstacle Avoidance Logic"}:::onboard
            Event_Packer["Event Serializer"]:::onboard
        end

        subgraph Actuators ["Action"]
            M_Propulsion["Motor Controllers"]:::action
        end
    end

    S_Vision & S_Telemetry --> Local_Avoidance
    Local_Avoidance --> M_Propulsion
    
    Local_Avoidance -. Telemetry_and_Decisions .-> Event_Packer
    Event_Packer ==>|State_Sync| G_ESKG
    
    G_Logic -. Strategic_Objectives .-> Local_Avoidance
    G_ESKG <--> G_Logic

Loading

Internal engine

At its core, Ringil is driven by continuous interaction between physics, local intent, and collective behavior.

$$\dot{p}_i = v_i \quad,\quad \dot{v}_i = u_i + f_i^{env}$$

Each node is first and foremost a physical system. Position evolves from velocity, and velocity evolves from applied control and environmental disturbances (wind, currents, drag, etc.).

$$u_i = -\nabla \left( \frac{1}{2}k_a |p_i - p_{goal}|^2 + \sum_{j} \frac{k_r}{|p_i - p_j|^2} \right)$$

This defines how an entity reacts with compromise between:

• moving toward an objective (attraction),
• and avoiding others or obstacles (repulsion).

$$\dot{p}_i = \sum_{j \in \mathcal{N}_i} a_{ij}(p_j - p_i)$$

There is no explicit formation controller here. Instead, each node continuously adjusts itself relative to its neighbors.

$$x_{t} = x_{t|t-1} + K_t \left(z_t - H x_{t|t-1}\right)$$