Signal Registry Architecture

Overview

The Signal Registry pattern provides clean separation between physics simulation and device protocol implementation. Physics engine operates on abstract signal paths, devices expose protocol-specific interfaces.

Key Components

ISignalSource Interface 

class ISignalSource {
    virtual std::optional<double> read_signal(const std::string& path) = 0;
    virtual void write_signal(const std::string& path, double value) = 0;
};

Contract: Path format is device_id/signal_id. Returns nullopt if signal unavailable.

SignalRegistry

Thread-safe cache implementing ISignalSource. Coordinates between physics ticker and device request handlers.

Key members:

  • signal_cache_ - Cached signal values (map<string, double>)
  • physics_driven_signals_ - Signals managed by physics (set<string>)
  • device_reader_ - Callback to read actual device state (for actuators)
  • registry_mutex_ - Protects all operations

SimPhysics

Physics engine depends only on ISignalSource*. Zero knowledge of ADPP protocol.

Signal flow:

  1. Read actuator states via signal_source_->read_signal()
  2. Evaluate signal graph (transforms, routing)
  3. Update physics models
  4. Write sensor values via signal_source_->write_signal()

Device Integration

Devices check registry before returning internal state:

// In read_signals():
std::string path = device_id + "/" + signal_id;
if (g_signal_registry && g_signal_registry->is_physics_driven(path)) {
    auto val = g_signal_registry->read_signal(path);
    if (val) return *val;
}
return internal_state[signal_id];

Thread Safety

Lock policy:

  • Physics ticker: Acquire registry mutex ā†’ read actuators ā†’ compute ā†’ write sensors ā†’ release
  • Device handlers: Acquire registry mutex ā†’ read values ā†’ release
  • No lock held during physics model evaluation (only during registry I/O)

Deadlock prevention: No nested lock acquisition. Physics and device mutexes are separate.

Signal Routing

In mode=sim, routing is declarative in the FluxGraph graph file referenced by simulation.physics_config (provider config only carries the path).

edges:
  - source: device_a/actuator # Read from device
    target: model_b/input # Write to model
    transform: { type: linear, scale: 100.0 }

  - source: model_b/output # Read from model
    target: device_c/sensor # Write to device (registry cache)
    transform: { type: first_order_lag, tau_s: 2.0 }

FluxGraph processes this schema; provider-sim does not define or validate it. See FluxGraph docs for authoritative schema details (docs/schema-yaml.md in the FluxGraph repository).

Device reads from the registry return physics-computed values when the signal is marked physics-driven.

Extension Points

Adding graph transforms/models/rules: update the FluxGraph graph and FluxGraph runtime as needed.

Adding devices: include registry checks in device read_signals() implementations.

Benefits

  • Protocol independence - Physics has zero ADPP dependencies
  • Reusability - Same physics engine works with any protocol via ISignalSource
  • Declarative config - Full signal flow visible in YAML
  • Testability - Mock ISignalSource for unit tests
  • Debuggability - Registry can log all signal transactions

Design Trade-offs

Cost: Additional indirection layer, mutex overhead on signal access

Benefit: Clean architecture boundaries, long-term maintainability, future-proof for signal recording/replay features