A state lattice planner searches over a pre-built set of feasible motion primitives arranged in a regular grid of states — giving vehicles drivable, high-quality paths by combining A*-style search with realistic motions.
A state lattice planner works from a fixed menu of realistic little moves the vehicle can make, laid out so they connect neatly into a grid of reachable states. It then searches that grid for a route — every path it finds is made of drivable moves.
🎯 Quick challenge
A state lattice planner guarantees drivable paths by…
For a car-like robot, a path is only useful if the vehicle can actually drive it. The state lattice planner guarantees that by searching over a fixed menu of realistic, drivable moves arranged into a neat grid.
The idea
A state lattice planner precomputes a set of motion primitives — short, kinematically feasible motions the vehicle can execute (gentle turns, straights, forward and reverse) — that are carefully designed to connect states in a regular lattice (a discretized grid of positions and headings). Because the primitives snap neatly onto lattice states, they can be chained together, and you can run a graph search (A*) over the lattice to find an optimal route. Every edge in that graph is a real drivable move, so any path the search returns is feasible by construction.
Search a grid of drivable moves
Because the building blocks are pre-verified drivable motions that connect lattice states, chaining them yields a path the vehicle can actually follow.
State lattice vs Hybrid A*
Both produce feasible paths for nonholonomic vehicles, and they're close cousins:
State lattice — uses a fixed, precomputed set of primitives on a regular grid. The regularity makes search clean and the primitives can be optimized offline for quality, but it's constrained to the lattice resolution.
Hybrid A* — generates motions on the fly from the continuous vehicle state, so it's more flexible in exactly where it ends up, at some cost in regularity.
Both appear in Nav2's Smac planner family, chosen by application.
Strengths and limits
Feasible by design. No post-hoc smoothing needed to make paths drivable.
Optimizable primitives. Motions can be designed offline to be smooth and efficient.
Resolution trade-off. Finer lattices give better paths but a larger graph to search; coarser lattices are faster but blockier.
Best for structured motion — road-like driving, parking, and defined maneuver sets.
Why it matters
The state lattice planner brings the power of grid-based search (A*) to vehicles that can't move arbitrarily — by making the search operate over feasible motions rather than idealized grid steps. It's a key technique for autonomous driving and constrained mobile robots, producing optimal and drivable paths.
