A useful starting point is to separate two things:
- The operating environment: depots are active workplaces, with people present for cleaning, daily checks, servicing, charging, and ad-hoc tasks. In practice, you rarely get a guarantee that an autonomous bus will always have a perfectly “cleared” zone.
- The safety architecture: where the system’s “eyes and judgement” live, and what dependencies exist outside the vehicle.
Smartbuses (depot-capable autonomous buses) take a vehicle-based approach: they aim to make each bus safe by design in mixed depot conditions.
Depot safety: people, big vehicles, and occlusions
Depots have a specific safety challenge that is easy to underestimate if you come from passenger cars: occlusion.
Buses are tall, long, and often parked in dense rows. People can be hidden behind vehicles, appear between bays, or step into a path with minimal warning. This is exactly why “blind spots” became such a serious safety topic in heavy vehicles in the first place – and why Europe has been tightening requirements for safety systems that protect vulnerable road users through the EU’s General Safety Regulation framework.
So the practical depot question becomes:
Can the autonomous system reliably detect people around a large vehicle, even when vehicles block lines of sight?
Two approaches you’ll see – and what they imply for safety
Depot autonomy concepts fall into one of two buckets:
1) Infrastructure-based perception and control
This approach instruments the depot with sensors on poles/buildings and uses a control system to coordinate movements. It can work, but its safety case tends to rely on additional assumptions:
- Coverage: sensors must “see” all relevant areas all the time.
- No gaps: blind zones can appear due to occlusion, layout changes, maintenance issues, weather effects on sensors, or simply the geometry of large vehicles.
In other words: the safety envelope is partly “inside the depot.”
2) Vehicle-based perception (smartbus approach)
This approach places the core perception on the bus itself – typically designed for 360-degree awareness around the vehicle – so the bus can operate safely in the space immediately around it without requiring the depot to be densely instrumented.
This shifts the safety logic:
- The bus carries its own “visibility bubble.”
- The most safety-critical detection problem (people near the vehicle) is handled locally.
- Infrastructure can still add layers (rules, geofencing, access control), but it’s not the primary source of situational awareness.
In other words: the safety envelope is primarily “inside the vehicle.”
How the architecture changes the safety dependencies
Different architectures place safety responsibilities in different places. From an operator risk perspective, there is a meaningful difference between:
- A bus that remains safe when parts of the depot system degrade, versus
- A bus whose safety depends on the depot system being fully available and perfectly maintained.
Depots change constantly: parking patterns evolve, temporary works appear, charging layouts shift, equipment moves. When safety depends on infrastructure coverage, every change creates a new question: “did we preserve visibility everywhere that matters?”
Vehicle-based perception is not “maintenance free” either – but its safety-critical sensing moves with the vehicle and is validated as part of the vehicle system, rather than being distributed across the site.
What “safe” looks like in practice for smartbuses
In depot autonomy, safety is primarily driven by predictable behaviour and defined operational rules:
- low, controlled speed
- predictable trajectories
- priority for pedestrians
- immediate stop behaviour under uncertainty
- defined operating boundaries (where autonomy is allowed, and where it is not)
- well-defined staff procedures (who supervises, how to resume after a stop, how staff interact with the vehicle)
In practice, safety is achieved through the combination of system behaviour, operating rules, and staff training.