By Syntell senior business development engineer Ziyaad Lee

For decades, a traffic signal in South Africa has been judged by one question: is it working? A controller ran a fixed time-of-day plan, a technician drove out when it failed, and success meant red, amber and green in the right order. That model kept the country moving, but it was never designed for today’s traffic volumes, or for the conditions under which road authorities now operate.

The industry is moving from managing intersections to managing networks, and from asking whether signals are on to asking how well the network performs. For an economy that depends on moving people, goods and minerals efficiently, that shift matters well beyond traffic engineering.

Intelligence layered onto an unreliable network does not produce better traffic; it produces confident, bad decisions. Adaptive control depends on accurate detection, dependable communications and controllers that stay online.

In South Africa, power instability, cable theft, vandalism and declining in-house technical capacity are engineering inputs, not footnotes, and a dark intersection is a dangerous one. Battery backup, hardened cabinets, remote fault detection and distributed architectures that reduce copper cabling make everything above them possible.

Timing plans were traditionally built from periodic traffic counts and revised every few years. Today, detectors, CCTV, enforcement cameras and GPS probe data describe corridor performance minute by minute.

Performance can now be measured in travel time, delay and reliability rather than equipment uptime alone. Continuous data also reveals what periodic audits miss. In Stellenbosch, cycle-by-cycle analysis exposed an all-red clearance interval at R44/Techno Avenue that was too short for measured approach speeds, which the municipality then corrected. This is how data stops being a by-product and becomes the basis for decisions.

A Connected Network

Signal control has long been treated separately from public transport, incident response, enforcement and freight. The real gains come when these are connected, so that a late bus, an approaching ambulance or an incident on a parallel route can inform how the network responds.

That depends on open, backwards-compatible technology. Most authorities run mixed-generation networks built over decades, so the realistic path is progressive modernisation. Start with the corridors where congestion, safety risk or outages are worse, prove the benefits, then expand, without stranding existing assets.

With reliable data and connectivity in place, AI begins to earn its keep. The Quebit adaptive optimisation platform, developed by Bytefuse AI in collaboration with Syntell, was connected to existing Syntell MX controllers on the R44 corridor in Stellenbosch and the Beach Road and York Street corridor in George. No controller was replaced, no firmware was modified and no civil works were needed: the AI layer communicates through the open UTMC protocol and uses FLIR TrafiCam AI cameras already in place.

A/B-tested results showed delay reductions of 53% to 57% in Stellenbosch and 30% to 50% in George. On the R44, median corridor delay in the afternoon peak fell from two minutes and 20 seconds to 56 seconds. Stellenbosch Municipality has since signed a 12-month service level agreement to expand the programme to 36 intersections.

This is not a choice between AI and rule-based control. The MX controller keeps authority over safety interlocks and phase sequencing, and reverts automatically to its stored plans if communication is lost. AI adds prediction and optimisation on top, and traffic engineers remain in charge.

Procurement has not kept pace with the technology. Traffic management systems are still largely bought as capital assets, yet software and connected infrastructure do not stand still. Self-hosted platforms tend to fall behind as upgrades and security patches wait for renewed budget cycles.

Service-based models tie payment to availability, support and continuous improvement, much as authorities already buy telecoms and office software. Concerns about operating budgets and multi-year commitments are legitimate, but they are better addressed through multi-year budgeting and clear service levels than by owning systems that quietly fall behind. The question is not whether to own or rent, but how to pay for networks that keep performing.

The Road Ahead

The same logic applies to mines, ports, logistics hubs and toll roads, where every minute of queuing at a constrained access point has a measurable price.

Over the next decade, connected vehicles, edge computing and real-time analytics will move traffic management from reacting to congestion towards anticipating it. For African cities, nearer-term autonomy is likely to sit in the network rather than the vehicle: systems that monitor, diagnose and optimise themselves.

Infrastructure procured today should be specified with that future in mind: open, interoperable and supported throughout its life. T

he opportunity is not to build more, but to make what we already have smarter, more connected and more resilient.