Where AI and simulation can improve transport decisions

Connecting the network is progress

A command centre is the operational hub of a transport network. It is where operators monitor conditions, respond to incidents and manage controls across the network, from traffic signals to lane controls and traveller information.

Many command centres still rely on ageing systems that were not designed to work together. Operators may have to move between legacy signal networks and separate camera platforms, while incident feeds sit elsewhere. The result is a fragmented view of the network at the very moment operators need a clear one.

That fragmentation leaves operators piecing together information across several screens, often while an incident is already unfolding. Bringing those feeds into one real-time view gives them a clearer picture of the network and helps them spot problems earlier.

But visibility only gets you so far. Knowing what is happening does not tell an operator which response will work best or, just as importantly, how that choice will affect the rest of the network. This is the next problem to solve. Avoidable congestion and road trauma cost the Australian economy tens of billions of dollars each year. AI and simulation can help operators look ahead, test possible responses and make a more informed decision before they act.

The unglamorous work behind every good decision

Before operators can rely on AI to look ahead, the data beneath it needs to tell a consistent story. Sensors can report at different intervals and systems may map the same network in different ways. Something as basic as an intersection can mean something different from one platform to another.

Making that data consistent takes time. Operators need detailed readings translated into a network-wide view they can actually use. It is not the part that usually makes the demonstration, but it is essential. If the underlying data is inconsistent, AI will work from an incomplete picture of the network. Its predictions may appear precise, but they can still lead operators towards the wrong decision.

From what's happening to what happens next

Once operators have a trusted view of the network, they can look beyond what is happening now. Prediction can show where pressure is likely to build and then simulation can test how the network may respond before an operator acts.

For example, an operator could compare the likely effect of adjusting a signal with closing a lane during an incident. That gives the command centre a chance to choose its response while there is still time to influence the outcome, rather than explaining afterwards why it went wrong.

But the best response is not always the one that moves traffic fastest. An operator may accept a longer average journey time to keep a corridor clear for emergency vehicles. In another situation, they may spread delays across two routes so one community does not carry the full impact. A useful model needs to understand why operators make those choices. It should support the priorities they are balancing in the control room, rather than optimise for the outcome that is easiest to calculate.

Value comes from changing a decision

This is where technology programs can lose sight of the outcome. Agencies do not need to improve every operational decision at once. Similar to the examples mentioned above, they should start with one or two decisions that have the greatest effect on the network and where a better choice can produce a clear, measurable result.

Traffic engineers are rightly cautious about relying on data and AI in an operational setting. Trust has to be earned by comparing the advice with what actually happened and being transparent about where confidence is high and where it is not. This gives engineers evidence of whether it improves the decision, how much weight to give the advice and when their own judgement should take precedence.

Once agencies can demonstrate better outcomes on those priority decisions, they have a stronger basis for introducing the capability into their command centres. Bringing the network together provides the foundation, but its value is realised when operators can use their experience and human judgment alongside technology to make a better decision in the moment. The question is not how to improve every decision, but which decision to improve first.