Motion Planning

The mental shift that makes motion planning tractable: stop thinking about the robot as a shape moving through the world. Think of it as a point moving through a higher-dimensional space.

~13 min

The classics, and the reason D*-lite still runs on real robots in 2026. Path-planning fundamentals you'll re-implement for 80% of mobile-robot work.

~14 min

Probabilistic roadmaps — precomputed paths for repeated queries in high-DOF spaces. The first sampling-based planner, still the right answer when you'll plan from many starts to many goals on the same map.

~12 min

The go-to sampling-based planner for high-DOF arms. How the tree grows, why RRT* gives asymptotically optimal paths, and the variants (informed-RRT*, BIT*) that made it ship in production.

~17 min

Planning via continuous optimization. The backbone of modern MPC, long-horizon arm motion, and aggressive drone trajectories. Direct collocation, shooting, and the practical patterns that converge.

~14 min

Gradient-based and sampling-based trajectory optimizers that power MoveIt's advanced planners. The two ways to refine a noisy initial trajectory into smooth, collision-free motion.

~12 min

When the line between control and planning blurs. MPC re-plans every tick over a finite horizon, replacing a slow planner + fast controller with a single layer that does both. The pattern autonomous cars adopted.

~12 min

Combining discrete logic (what to do) with continuous motion (how to move). The frontier of long-horizon autonomy — and the architecture behind 'put the dish in the dishwasher.'

~13 min

The structure every modern robot-autonomy stack (Nav2, NASA, Sony) converged on. Why trees beat state machines, and how to actually use BT.cpp / py_trees in production.

~13 min

POMDPs, belief-space planning, and the algorithms for robots that can't see everything. Why most robot tasks are partially observable, and the practical methods that handle it.

~13 min