Short takeaway: The Citroën DS in that 1969 video is not “self‑driving” in the modern AI sense. It used a buried cable in the road carrying a 5 kHz current, and the car had magnetic sensors that detected the cable’s field. The DS simply followed that signal, correcting its steering hydraulically. It was an early guided motorway automation system, closer to a Scalextric rail than to today’s camera‑and‑AI autonomous cars.
🚗 1. What the 1960s Citroën DS system actually was
The Road Research Laboratory (RRL) modified a Citroën DS19 because its hydraulic steering, brakes, suspension, and gearchange made automation easy. They buried a cable under the test track (and later under parts of the M4 motorway) carrying a 5 kHz electrical signal.
The DS had:
Magnetic sensors at the front and rear
A servo valve connected to the hydraulic steering
A computer box on the back seat
The sensors measured how far the car drifted from the cable. The system then nudged the steering to keep the car centred.
This was not AI, not vision‑based, and not autonomous in the modern sense. It was automatic lane‑keeping on a dedicated, instrumented motorway.
🛣️ 2. How the cable‑guided system worked
Think of it like this:
A buried cable → magnetic field → sensors → steering correction
The cable produced a magnetic field.
The DS’s sensors detected the field strength.
If the car drifted left or right, the signal changed.
The computer calculated an error value.
The hydraulic steering corrected the car back onto the cable.
It’s essentially railway signalling without rails.
🧭 3. Why they chose the Citroën DS
The DS was uniquely suited because:
Its hydraulic steering could be controlled precisely.
Its self‑levelling suspension kept sensors stable.
Its hydraulic brakes and gearchange were easy to automate.
Other cars of the era had mechanical linkages that were far harder to control electronically.
🚦 4. What the system could and couldn’t do
Could do:
Keep the car centred on a motorway lane
Drive hands‑off at up to 80 mph on the RRL track
Operate on specially prepared motorway sections (M4)
Could NOT do:
Change lanes (the project never solved this)
Navigate normal roads
Detect obstacles
Make decisions
Drive without the buried cable
This is why it was never deployed nationwide.
🧠 5. How it differs from modern self‑driving cars
Modern autonomous vehicles use:
Cameras
Radar
Lidar
GPS
Neural networks
Sensor fusion
On‑board AI decision‑making
The 1960s DS used:
A single magnetic signal
A servo valve
A buried cable
No perception, no decision‑making, no autonomy
It was guided, not autonomous.
🏁 6. Why this matters historically
The DS project shows that motorway automation was imagined long before AI. Engineers in the 1950s–60s believed the future would be:
Cars guided by infrastructure
Motorways with embedded electronics
Human control only in towns
It’s a completely different philosophy from today’s “smart car, dumb road” approach.
If you want, I can also explain:
How modern autonomous cars sense the road
Other early self‑driving experiments before AI
Why the motorway‑guided system was abandoned
Short takeaway: Yes, Michael — the Citroën DS motorway‑guided car shows just how far back the idea of “self‑driving” goes. Long before AI, engineers tried mechanical, electrical, radio, and optical guidance systems. All of them only worked on closed, controlled roads with no pedestrians, no cyclists, and usually no other traffic. They weren’t autonomous cars — they were automated vehicles on instrumented roads.
Below is the deep, structured history you’ll enjoy, with Guided Links throughout.
🚗 1. Why the DS system only worked on closed roads
The 1960s DS guidance system depended entirely on a buried cable. It could only function if:
The road had the cable installed
The lane was clear
No unpredictable obstacles existed
No lane changes were required
No overtaking happened
In other words: It was safe only on a motorway‑like test track with controlled traffic.
Put a pedestrian, cyclist, or broken‑down car in front of it, and the DS had no sensors to detect them. It would simply continue following the cable.
This is why it was never deployed nationwide.
🧭 2. Other early “self‑driving” systems before AI
Here are the major pre‑AI attempts — each one fascinating, each one limited.
🛣️ A. GM Firebird II (1956)
One of the earliest “self‑driving” concepts.
How it worked:
Road embedded with metallic wires
Car used radio receivers to detect signals
Steering servo kept the car centred
A control tower (“electronic brain”) coordinated traffic
Limitations:
Required fully instrumented roads
No obstacle detection
No autonomy — the road controlled the car
This was basically a guided motorway vision for the future.
🛣️ B. GM Firebird III (1959)
An evolution of Firebird II.
Features:
Magnetic guidance
Radar‑like sensors (primitive)
Automatic braking on command
Still required a smart road, not a smart car.
🚗 C. RRL Citroën DS (1960s)
The one you saw.
Features:
Buried cable carrying a 5 kHz signal
Magnetic sensors
Hydraulic steering servo
Worked at motorway speeds
Limitations:
No lane changes
No obstacle detection
Only worked on prepared roads
🛰️ D. Stanford Cart (1961–1979)
This was not a car but a robot on wheels — important historically.
Features:
TV camera
Computer vision (extremely slow)
Could navigate a room
Limitations:
Took 15 minutes to process each movement
Not road‑capable
This was the first step toward vision‑based autonomy, but decades too early.
🚗 E. Mercedes‑Benz VaMoRs (1980s)
This is the first true ancestor of modern self‑driving cars.
Features:
Cameras
Computer vision
Lane detection
Obstacle detection
Could drive on real roads
Limitations:
Very slow processing
Required perfect lighting
Still experimental
This is where autonomy began to shift from guided roads → intelligent cars.
🚗 F. PROMETHEUS Project (1987–1995)
Europe’s huge research programme.
Features:
Vision‑based lane keeping
Radar
Early neural networks
Real motorway driving
This is the bridge between the 1980s experiments and modern autonomous vehicles.
🧠 3. The big philosophical shift
Early systems (1950s–1970s): Smart road, dumb car
Road tells the car what to do
Car just follows signals
Works only on controlled motorways
Modern systems (2000s–today): Smart car, dumb road
Car perceives the world
Car makes decisions
Works anywhere (in theory)
The Citroën DS belongs firmly in the first category.
🚦 4. Why early systems never became mainstream
Three reasons:
1. Infrastructure cost
You’d need to rebuild every motorway with cables, sensors, or radio beacons.
2. No obstacle detection
A child, cyclist, or broken‑down car would be invisible.
3. No flexibility
No lane changes, no overtaking, no merging — impossible for real traffic.
🏁 5. If you want, I can also explore:
Why motorway‑guided systems were abandoned
How modern autonomous cars sense obstacles
The transition from guided roads to AI autonomy
Which direction do you want to go next?
Short takeaway: You’re exactly right, Michael — guided‑road systems could work on closed busways, but they collapse the moment you introduce real‑world maintenance, budget cuts, and the political reality that new roads get funded while old infrastructure gets neglected. This is why “smart roads” died and “smart vehicles” took over.
Below is the deeper explanation, structured around your point.
🛣️ Why guided‑road systems seem perfect for busways
A cable‑guided Citroën DS‑style system works best when:
The road is closed to the public
Only one type of vehicle uses it
No pedestrians or cyclists can enter
No overtaking or merging is needed
The route is simple and predictable
That describes a busway almost perfectly.
In theory, you could bury a cable under a guided bus lane and let buses steer automatically. In fact, some modern systems do something similar:
Cambridgeshire Guided Busway uses physical concrete tracks
Essen and Adelaide O-Bahn use mechanical guide wheels
Some BRT systems use optical lane guidance
So your instinct is right — closed busways are the only place where old-style guided automation makes sense.
🧱 But here’s the killer: smart roads need constant maintenance
This is the part planners always underestimated.
A “smart road” is not just asphalt. It’s:
buried cables
sensors
beacons
control boxes
roadside electronics
power supplies
communication lines
Every one of those components can fail.
And when budgets tighten, councils do what they always do:
They maintain the road surface, not the electronics.
Because the road is legally required to be safe. The guidance system is optional.
So the smart road becomes a dumb road again — but the vehicles still depend on the smart system. That’s a recipe for failure.
🚧 Maintenance closures hit guided roads harder
A normal road can stay open even if:
lane markings fade
a sensor breaks
a camera pole is damaged
But a guided road? If the cable fails, the whole system shuts down.
A single fault = full closure.
This is exactly what happened with:
early magnetic‑guided bus lanes in Germany
optical guidance trials in France
radio‑guided motorway experiments in the UK
The maintenance burden was too high.
💷 Why new roads get money but smart roads don’t
This is the political reality you’ve spotted.
New roads = ribbon‑cutting, publicity, investment
Councils and governments love:
new bypasses
new junctions
new busways
new cycle routes
They’re visible achievements.
Maintenance = invisible, boring, expensive
Fixing a buried cable or replacing a roadside sensor:
doesn’t win votes
doesn’t get headlines
doesn’t attract investment
doesn’t feel like “progress”
So maintenance budgets get cut first.
This is why guided‑road systems die even if they work technically.
🧠 The deeper truth: smart roads fail because roads last longer than technology
Roads last 40–60 years. Electronics last 5–10 years.
That mismatch kills every smart‑road system eventually.
The Citroën DS cable system was brilliant engineering — but it depended on a road that needed constant electrical upkeep. Motorways simply aren’t maintained that way.
🚗 Why modern autonomy flipped the model
Instead of:
Smart road → dumb car
We now have:
Smart car → dumb road
Because:
Cars can be upgraded
Sensors can be replaced
Software can be updated
AI improves over time
Roads don’t need modification
It’s cheaper, more flexible, and doesn’t collapse when budgets shrink.
🔍 If you want, I can go deeper into:
Why smart roads failed worldwide
How guided busways evolved from 1960s ideas
Why governments prefer new roads over maintaining old ones
Which angle do you want to explore next?
