Naturalistic Data Flashcards
What is Naturalistic Data?
Naturalistic Data
Naturalistic data is big data collected in real-traffic, by road users performing their usual daily activities. Traditionally recorded from instrumented cars and trucks.
3 Vs of Big Data.
Big Data 3Vs
- Volume
- Velocity
- Variety
4Ms of Naturalistic Data
Naturalistic Data 4Ms
- Much
- Many
- Missing
- Messy
Explain Development and evaluation of active safety systems
Explain Field data.
Field Data
Field data is data collected in real-traffic often according to an experimental protocol, which may define driver type (e.g. test engineer) and driving conditions (e.g. itinerary). It may require instrumented cars/trucks and infrastructures.
Explain FOT.
Field Operational Test (FOT)
Field operational test are large collections of real-traffic data, often performed in a naturalistic fashion, but with the aim of evaluting a system (e.g. an active safety system). As a consequence, a treatment and a baseline phase are often present in an FOT.
Naturalistic Studies, Field Operational Tests and Real-world Tests correlate?
What Is Naturalistic Data Good for?
Explain Heinrich’s Triangle
Distribution of crashes and outcomes?
What does Henreich’s Triangle state?
What can be used as surrogate for Crashes?
Near Crashes as Surrogate for Crashes
- Why would you ever use near crashes?
- Because
- crashes in naturalistic datasets are just a few
- crashes in naturalistic datasets tend to be mild anyway (so we still need to climb the Heinrich’s triangle)
- if near-crashes are indeed a good surrogate for crashes, it may not be ethically correct to ”wait” for crashes to be recorded.
How can Naturalistic Data help in addressing som major HF challenges?
Can help address some of the major Human Factor challenges:
- Transitions: when, how, what, (why) is appropriate
- HMI: mode confusion, support
- Interaction with other non-autonomous road users
- Learning: adaptation, habituation, and new behaviours
- Network simulations and models: interaction of the different road-user interractions, mixed traffic, etc…
- Where does human-machine cooperation in autonomous drivign work and where does it fail (and why)?
What does naturalistic data offer?
What nat data offers
- Naturalistic environment
- Long term (adaptation, learning, new strategies, etc.)
- Wider population (demographics, styles, etc.)
- Different driver status
- Different driving contexts
- Network effect (several road users interacting; mixed traffic)
Explain Odds Ratio
Odds Ratios
- Analysis methodology from epidemilogy.
- Assess the association between the absence or presence of two properties.
- Because there are 2 properties and can have 2 states (present and non present) all possible combinations are representable in a contingency table with 2 rows and 2 columns:
Interpretation:
- With baseline: If you use the phone, your odds to crash are double.
- With near crashes: If you use the phone, your odds to crash are double as much as experiencing a near-crash.
- But is this statistical significant?
How to calculate CI?
Confindence Interval
When is CI not significant?
