“Driving is not merely the mechanical act of lane keeping; it is the continuous maintenance of an internal simulation of the world.”
It feels somewhat ironic that human motorists can sharpen their driving craft by studying autonomous driving systems. Yet, looking through that lens is surprisingly illuminating. When engineers set out to teach a machine how to navigate real-world traffic, they are forced to deconstruct intuition and model every layer of perception, prediction, and control into a formal theoretical architecture. Those mathematical abstractions might not capture every subtle nuance of human instinct, but they provide an invaluable conceptual framework for understanding what our brains actually do behind the wheel.
Driving Agent
- The Agent: The driver paired with the vehicle platform.
- The Environment: The roadway geometry, surface friction, weather conditions, dynamic road users, and static traffic infrastructure.
- The Action Space: Steering angle rate, throttle input, braking pressure, gear selection, and active signaling (indicators, horn, lane positioning).
Crucially, driving is not a fully observable board game like chess; it is a Partially Observable Markov Decision Process (POMDP). You never have access to the complete ground truth of the world. Other road users have occluded sightlines, hidden intentions, and unpredictable reaction times. To bridge the gap between noisy sensory input and safe vehicle control, the driver must continuously build test, and update an internal belief state—a dynamic mental model of the immediate world.
Observation and Anticipation
Observation is the passive intake and active scanning of environmental features: checking the rear-view mirror before braking, glancing down a blind side street, noting a wet patch of tarmac, or scanning the body language of a pedestrian poised near the kerb. It tells you “What is the state of the world right now?”
Observation without anticipation is merely passive sightseeing. A camera records data; a driver projects it forward in time. Anticipation is a rolling forward simulation (akin to a Monte Carlo rollout or trajectory planning algorithm). When you see a delivery van double-parked 100 metres ahead on a narrow residential street, anticipation simulates the probable downstream effects before they physically manifest:
- Will an oncoming vehicle arrive at the bottleneck before me?
- Is a courier about to swing the driver-side door open?
- Will a pedestrian emerge blindly from behind the front bumper?
It tells you “Where will the conflict points be in next few seconds?”
When a novice driver sits behind the wheel, their observation–anticipation loop runs primarily through deliberate, high-effort cognition (System 2). They must consciously remind themselves to look at the limit point of a bend, deliberately calculate whether their alloy wheels will clip a kerb, and actively ponder when to lift off the accelerator. The process is high-latency, mentally exhausting, and easily overwhelmed by sudden events. With deliberate practice, this loop is compiled down into fast, low-latency heuristics (System 1):
- Spatial Anchoring: You intuitively perceive the physical corners of the car without conscious spatial triangulation.
- Micro-Cues: A slight twitch of a front tyre on a stationary vehicle at a side road registers instantly as a risk of them pulling out.
- Temporal Buffering: Your foot automatically covers the brake pedal before your conscious mind articulates why. In neurological terms, this sensory processing, belief updating, and motor planning executes in under 200 milliseconds.
Before AI take over
Modern autonomous driving systems are packed with cameras, radars, and LiDARs to construct a holistic 360-degree view, supported by billions of dollars poured into predictive modelling and automated decision-making. Yet, ironically, they burn orders of magnitude more energy than you just to make a single manoeuvre.
In the near future, driving will likely be dominated by autonomous systems—it feels like an inevitable tide. On balance, it’s a positive shift: roads will be safer, human fatigue will be spared, and perhaps only driving schools stand to lose everything.
But we aren’t there yet. Even behind the wheel of a car with the latest autonomous tech, your eyes and mind still need to be locked on the road and the surrounding environment. And just like any sport, learning and practicing this decision-making loop over and over is how you build your true decision muscles.