What is going on at Toyota’s automated driving programs? Sagar Behere has the answer! He works as Manager AD Architecture and Functional Safety at Toyota Research Institute, USA and will present his Key Note Address Cybersecurity for Highly Automated Driving in Track E Cybersecurity & Connectivity during the second day of Vehicle Electronics & Connected Services. Sagar gives us a sneak peak of his presentation and shares his thoughts on the big challenges for reaching full automation. Besides that, he is providing us with a quote from Isaac Asimov…
Could you briefly explain your role, involvement and experience in leading development of AD architecture and functional safety at the Toyota Research Institute in California?
– System Architecture and Functional Safety are very closely inter-related. This is especially true for large, complex, and safety-critical systems. For such systems, it possible to develop many feasible solutions, each of which may have differing impacts on safety. Within this solution space, making the right decisions requires simultaneous consideration of architecture and safety. My role at TRI is all about facilitating such decisions to create high performing and safe automated driving systems. More information about Toyota’s automated driving programs and philosophy can be found at http://automatedtoyota.com/
What do you consider as the biggest challenge for reaching full automation?
– I think there are challenges all across the board. Technically, a big challenge lies in developing systems that can provably safely handle the vast variety of edge cases, each of which are individually rare but collectively more common. Handling the edge cases requires deep semantic understanding of the situation. There is a very big difference between sensing the state of the world through various modalities and actually understanding its semantics sufficiently to continuously take the correct actions. As humans, we rely tremendously on our ability to interpret what a situation _means_ in terms of the actions we should be taking. That meaning is constructed from much, much more than the data our senses send to our brain.
According to you, when will we reach full automation (level 5)?
– Heh, I am going to revert to a quote from my favourite science fiction author, Isaac Asimov: “There is as yet insufficient data for a meaningful answer.”
Which competences will be most crucial for a successful development of fully autonomous vehicles?
– In terms of technical capabilities, I think that a necessary competence is the prediction of how the world will evolve, especially in the presence of various possible actions of involved actors, and then planning in the presence of multiple hypotheses. Apart from that, there have been significant improvements in machine learning during the last decade. However, it remains challenging to make meaningful, provable statements about the safety of machine learning systems. This is in fact an active area of research. Beyond technical competencies, highly capable autonomous vehicles are likely to be among the most complex machines ever built. Developing these as products, in rapid cycles and with variants, will push the boundaries of how we do systems engineering (Integration, Verification, Validation, Qualification).
What are you most looking forward to by attending and presenting at VECS 2018?
– The program seems to be both broad and deep. I look forward to interesting insigts and engaging conversations with colleagues in the field.
Without revealing too much, what will you cover in your presentation Cybersecurity for Highly Automated Driving at VECS 2018?
– There has been quite a buzz about cybersecurity and vehicles in recent media. Hollywood movies too! In my talk, we will keep aside the buzz and take a hard look at the real technical problems, solutions, and challenges in the field.