ASVNavigationStack
Planning-Stage Autonomy Architecture for Surface Vessels
A planning-stage navigation systems project for autonomous surface vessels at SeaPower, focused on how route planning, obstacle avoidance, state estimation, and guidance logic should interface with vessel control in real marine conditions.
What the Project Is Actually Trying to Define
This is not presented as a finished navigation stack. The current work is about defining the right layer boundaries so future implementation effort lands on a coherent architecture.
Path planning
How the vessel should reason about route-level intent, waypoint structure, and mission progress before local maneuvers begin.
Obstacle avoidance
How local reactions should sit on top of route intent without turning the navigation layer into a reactive-only controller.
State estimation
How sensor fusion should provide navigation-grade state, not just raw telemetry, under drift, noise, and marine disturbances.
Control interface
How guidance and path-following outputs should be expressed so vessel-level controllers can act on them safely and predictably.
A Planning Workflow Before an Implementation Roadmap
The proposed stack is organized around the questions the vessel must answer, from mission intent through guidance and control handoff.
The Hard Part Is the Interfaces
What exactly does the navigation layer hand to vessel control?
The interface needs to decide whether control consumes waypoints, headings, surge/yaw references, path-parameter targets, or another guidance primitive.
How much marine behavior belongs in navigation versus control?
Current, drift, docking maneuvers, and low-speed handling can be modeled in either layer. The wrong boundary makes both layers harder to reason about.
When should local avoidance override mission intent?
The stack needs explicit arbitration between route commitment, collision avoidance, and operator intervention instead of assuming one objective always wins.
What happens when state confidence drops?
Fallback logic, degraded modes, and operator supervision need to be part of the architecture now rather than being treated as post-hoc exceptions later.
Why This Cannot Just Borrow a Generic Robotics Stack
Slow drift still matters
Marine platforms can be low-speed and slow to settle, but wind, current, and drift make small errors accumulate into operationally meaningful deviations.
Operator supervision stays in the loop
Even strong autonomy stacks need supervisory visibility, fallback logic, and operational handoff paths that fit real vessel operations.
Guidance is not the controller
For vessels, the path-intent layer and the low-level actuation layer can diverge sharply if the control interface is not defined with discipline.
What This Page Is and Is Not Claiming
- • The intended stack layers and system boundaries are being defined.
- • Marine-specific navigation questions are being made explicit early.
- • The project is tied to SeaPower-facing autonomy work, not a generic side exercise.
- • No deployed vessel-navigation stack is being implied here.
- • No sea-trial metrics or hardware performance claims are being made.
- • No fake implementation depth is used to make a planning-stage project look finished.