systems theory//OODA loop

The OODA loop is a model of decision-making as a cycle of four stages (observe, orient, decide, act), formulated by the US Air Force strategist John Boyd, and it is used in military doctrine, defence systems, incident response and operations to reason about one thing: who wins when two actors, or an actor and a changing world, keep reacting to each other. Boyd's claim is that the side that closes the loop with sufficient quality faster than the environment changes wins, because its decisions land on the world as it is, while the slower side keeps acting on a picture that has already moved.

The OODA loop is a model of decision-making as a cycle of four stages (observe, orient, decide, act), formulated by the US Air Force strategist John Boyd, and it is used in military doctrine, defence systems, incident response and operations to reason about one thing: who wins when two actors, or an actor and a changing world, keep reacting to each other. Boyd's claim is that the side that closes the loop with sufficient quality faster than the environment changes wins, because its decisions land on the world as it is, while the slower side keeps acting on a picture that has already moved.

It is the decision-level reading of a rule control engineers know from the inside: in any closed-loop system every stage eats part of a fixed time budget, and the delay between measuring and acting is what destabilizes a loop (delay as the enemy). A decision that arrives late is the right decision for a world that no longer exists. Orient is the expensive stage: it is where observations are fused with models, experience and expectations into a picture of the situation, the role a tracker and a multi-target tracking picture play in a defence system, and where most of the time and most of the errors live.

A good decision in time beats an optimal decision late. Engineering turns that into a stopping rule: if an anytime algorithm's plan improves as U(t)U(t)U(t) and every second of waiting costs ccc because the world moves, keep thinking while dU/dt>cdU/dt>cdU/dt>c and act as soon as the marginal gain falls below the cost of waiting.

In machines the loop is layered, each layer with its own deadline. A drone's attitude control runs at 1 kHz on the microcontroller, a robot's local planner at 10 to 20 Hz on its onboard computer, the global planner at about 1 Hz, and a fleet's task assignment every few seconds on a server; each must answer within its period, judged by its worst case, never its mean (hierarchical control).

The default answer is part of the design. When the decision layer misses its deadline, the controller keeps the last valid plan or falls back to a safe manoeuvre (stop, hold position), and designing that response matters as much as the algorithm (fail-safe design).

Against an opponent, speed is also a weapon. Getting inside the other side's loop, acting before it has oriented on your last move, is the core of Boyd's argument, and game theory is where an adversary's own choices enter the analysis.

It is a frame that comes with no guarantees. Unlike a feedback control loop, it has no stability test and no equations; it tells you where to look for lost time, and the tools that remove it (faster sensing, cheaper estimation, anytime planning, decisions pushed down to the layer that has the data) come from the rest of the loop.