Driverless but Not Directionless: The Positioning Infrastructure Australia Must Build Before Autonomous Vehicles Can Safely Take the Wheel

The popular narrative around autonomous vehicles tends to focus on the software — the neural networks, the sensor fusion algorithms, the ethical decision trees that govern split-second choices. What receives considerably less attention is the quieter, more foundational question: where, precisely, does the vehicle believe it is? And in Australia, that question does not yet have a consistently reliable answer.

Self-driving systems at the higher levels of automation do not merely need to know which suburb they are in. They need to know their position to within a few centimetres, in real time, with near-zero latency, and across the full range of environments Australian roads present — from dense inner-city corridors in Melbourne and Sydney to sun-bleached regional highways in Queensland and Western Australia. Delivering that level of positioning fidelity, continuously and without interruption, demands infrastructure that Australia is still in the process of building.

What Autonomous Systems Actually Demand From Positioning

A standard smartphone GPS operates with an accuracy of roughly three to five metres under favourable conditions. That is entirely adequate for turn-by-turn navigation. It is wholly inadequate for a vehicle travelling at 100 kilometres per hour that needs to maintain lane discipline, identify the edge of a shoulder, or execute a merge in moving traffic.

High-level autonomous vehicles rely on a layered positioning approach. Raw GNSS signals from constellations including GPS, Galileo, and BeiDou are augmented by real-time correction services — most commonly Real-Time Kinematic (RTK) or Precise Point Positioning (PPP) — to bring positional uncertainty down to the centimetre range. These corrections are delivered via ground-based reference networks that broadcast differential data to the vehicle, allowing onboard systems to resolve ambiguities in the satellite signal and compute a highly precise position fix.

The problem is that Australia's reference station network, while improving, is not uniformly dense. In metropolitan areas, the coverage provided by networks such as the state-based CORSnet systems is reasonably robust. Move beyond the urban fringe, and the density of reference stations thins considerably. Correction latency increases. Accuracy degrades. The centimetre becomes a decimetre, then approaches a metre — and at that point, the safety calculus for an autonomous vehicle changes fundamentally.

The Urban Corridor Problem

Even within Australian cities, the positioning challenge is more complex than infrastructure density alone suggests. Urban canyons — the corridors formed by tall buildings in CBDs — create multipath conditions in which satellite signals reflect off surfaces before reaching the vehicle's antenna. The receiver processes these reflected signals alongside the direct ones, introducing positional errors that correction networks cannot fully compensate for.

This is not a hypothetical concern. Studies conducted in dense urban environments globally have demonstrated that multipath-induced errors can push GNSS positions several metres from their true values, even when correction services are active. For a vehicle navigating a narrow CBD lane with tram tracks, cyclists, and pedestrians in close proximity, a multi-metre positional error is not a rounding problem — it is a collision risk.

Australian cities present their own particular complications. Melbourne's extensive tram network introduces both physical constraints and electromagnetic interference considerations. Sydney's varied topography creates localised signal occlusion. Brisbane's rapid urban densification means that the built environment is changing faster than mapping and infrastructure updates can track.

Regulation Is Catching Up, But Slowly

Australia's regulatory framework for autonomous vehicles has been evolving through the National Transport Commission and various state-level bodies. The Australian Road Rules have been progressively amended to accommodate automated driving systems, and trials have been conducted in multiple jurisdictions. However, the regulatory conversation has largely centred on liability, operational design domains, and safety standards — with the specific question of positioning infrastructure requirements receiving comparatively limited formal attention.

There is no nationally mandated minimum positioning accuracy standard for autonomous vehicles operating on public roads in Australia. Individual developers and fleet operators set their own internal thresholds, which vary considerably. This creates a situation in which a vehicle certified as safe under one operator's standards might be operating at the edge of acceptable accuracy in a corridor where another operator would not deploy at all.

The absence of a unified national positioning accuracy framework is not merely a bureaucratic gap. It is a structural vulnerability in the safety architecture that autonomous vehicles depend upon.

The Infrastructure Investment Question

Closing the positioning gap requires sustained, coordinated investment across several dimensions. The density of continuously operating reference station (CORS) networks needs to increase, particularly along major freight corridors and in regional centres that are increasingly being considered for autonomous vehicle applications — including last-mile logistics and agricultural transport.

High-definition mapping, which provides the prior spatial knowledge that autonomous systems use to contextualise their GNSS position, also requires continuous updating. A map that is six months old in a suburb undergoing significant development is a navigational liability. The fusion of real-time positioning with current HD map data is essential, and both components must be maintained to the same standard.

There is also the matter of resilience. Autonomous vehicles cannot be designed around the assumption that GNSS signals will always be available. Tunnels, underground car parks, and dense urban canyons all present environments where satellite visibility is severely degraded or entirely absent. Inertial navigation systems, LiDAR-based localisation, and visual odometry provide complementary positioning inputs, but integrating these with GNSS in a way that maintains centimetre-level accuracy through signal transitions remains an active area of research and development.

A National Opportunity Framed as a Technical Detail

It would be a mistake to treat Australia's positioning infrastructure deficit as a narrow technical matter best left to engineers and satellite geodesists. The deployment timeline for autonomous vehicles — and the safety record those vehicles will establish in their early years on Australian roads — will be shaped in part by decisions made now about where to invest in reference networks, how to regulate positioning standards, and how to coordinate between the federal government, state transport agencies, and private technology developers.

Australia has genuine strengths to build upon. Geoscience Australia's national geodetic infrastructure provides a rigorous datum foundation. State-based CORS networks have grown substantially over the past decade. The country's relatively low population density in many regions actually simplifies some aspects of the positioning problem, even as it creates coverage challenges in others.

The centimetre that separates a safe autonomous manoeuvre from a dangerous one is not simply a measure of distance. It is a measure of how seriously Australia treats the geospatial foundation beneath its most ambitious transport technology. Getting that foundation right is not optional — it is the precondition upon which everything else depends.