Many of us have already taken part in a heated debate on whether autonomous driving will be pervasive in the near future. Putting legal and other dimensions aside, I was (and still am) skeptical about the technical plausibility of autonomous driving. Because driving in reality is a more complicated endeavor than a historical extract of traffic situations on which autonomous driving is based. The question is not about whether autonomous driving works or not, it is about the complex (mixed human/machine) environment in which it operates -- that is made worse by unpredictable human behavior. If autonomous driving car would be deployed in an environment where no human actor is involved, it could perform its task perfectly. But in a mixed environment, I don't think it will be the case and direct conversion be a feasible.
What do you think are the major challenges that hinder autonomous driving from taking off?
As per my information and knowledge, till time all the deployed AVs have mixed versions (Human and Machine). So, I think problem is not how to deploy, but real issue is how to deploy it in all environments i.e. all roads, where conventional and autonomous vehicles both available for joy and ride.
Thank you.
Ankit R. Patel, by 'mixed environment' I was referring to the fact that the traffic is composed of AVs, vehicles driven by humans and those AV+human driven vehicles.
Thank you.
Sisay Chala In that case, you can consider the issues related with infrastructure, legal and ethical concerns, cyber attacks, government regulations and most important is how they implement it effectively, human mentality and habits to accept or not the vehicle once they switch over from manual to autonomous vehicles, etc.
Thank you.
Dear Sisay Chala,
from the experience in our research the problem of autonomous driving is not that we need and environment where we just have robots, it is more about that the current state of the art of algorithms is not capable of performing the full human cognitive power for driving a car. We see in each of the different areas of autonomous driving (perception, planning, control) always new edge cases which leads to importance of enhancing the algorithms. E.g. until today it is unclear which sensor setup is the best for "just" perceiving the environment. In addition in the field of AI we made a lot of progress but none of the algorithm is ready for vehicle integration when there is no holisitic safety assessment.
In my research e.g. we are focusing on autonomous racecars --> High velocities, High accelerations, non deterministic environment: We figured out a lot of problems in each of the fields of autonomous driving and therefore i think we still have a few years to go.
There are multiple human factors that must be considered in the transitional stage towards a fully automated road transportation system. A mixed traffic environment (semi-automated/fully automated vehicles alongside manually controlled vehicles) will make inter-vehicle communication a challenge. What cues/signals can be used to reliably estimate/infer driver intent, decision-making and behaviour? How consistent are these cues across the driving population and what factors must be taken into account (i.e. the problem of driver heterogeneity)? How are driver cognition and behaviour modulated by the environment (e.g. wet/foggy weather, day vs night, rural vs urban), as well as the LOA of their own vehicle /fellow vehicles and their trust-in-automation?
One potential agenda for human factors research/engineering is to explore more deeply driver cognition and behaviours relevant in mixed-traffic environments, and develop driver state monitoring systems (DSMS) that can track signals that facilitate cross-vehicle coordination. This will require significant research effort across several years.
Thank you Rohit Kenge for your input. Could you please elaborate a bit what you meant by "what to do in the critical sudden situations or incidents"?
Do you mean by actions that should be triggered when unexpected incidents occur such as, e.g., a sudden falling of objects from above?
- Badges
- Science topic