Nine questions / One connected system
Answers from the control loop
The boundaries between subsystems are where many of the useful questions begin. Open a question to follow the connection.
What makes a formula racing car autonomous?
An automated driving system performs the driving tasks within defined conditions, using sensing, decision-making and control. Autonomy does not remove the need for supervision arrangements, event rules or a safety process.
Does autonomous racing require artificial intelligence?
Machine learning can support perception and prediction, but estimation, optimisation, classical control and explicit safety logic remain important. There is no requirement for every subsystem to learn.
Why combine cameras, lidar and inertial measurements?
They provide different kinds of evidence and have different limitations. Fusion can improve state estimation when timing, calibration and uncertainty are handled correctly; adding sensors alone does not guarantee robustness.
Why not follow a pre-recorded racing line?
A stored line can be a reference. It cannot, by itself, account for a changed obstacle, a degraded sensor or a lower grip level. Feedback is what makes the reference usable under change.
What is the difference between planning and control?
Planning chooses a feasible intended motion over a horizon. Control converts that reference into actuator requests while correcting measured error. Their models and limits must agree even when their algorithms overlap.
Why does braking affect steering authority?
Longitudinal and lateral forces draw on the same tyres’ available capability. Strong braking can leave less capacity for cornering, so planning and control need to account for combined force demands.
Is a simulator enough to validate the car?
No. Simulation explores scenarios efficiently, but its models have limits. Hardware and physical tests are needed to examine timing, interfaces and behaviour that the simulator may not reproduce.
Does successful racing prove public-road readiness?
No. A closed circuit has different actors, rules and operating conditions from public roads. Evidence from one domain cannot establish safety in another without a separate argument and appropriate assessment.
Where should an engineering reader begin?
Begin with the perception essay and follow the sequence through planning, control and validation. At each interface, ask what is known, how old that information is and what happens when an assumption fails.