Thursday, July 14, 2016

Professor Gadh addresses issues relating to self-driving cars and artificial intelligence panel

Rajit Gadh, a professor at UCLA and one of the panelists on this evening’s panel discussion of self-driving cars and artificial intelligence at High Tech Los Angeles (HTLA), provided Etopia News with these comprehensive and detailed answers to three questions posed to him about these subjects:

1.      How feasible would it be to create a citywide network in Los Angeles that could manage the operations of millions of human-driven and autonomous vehicles using machine learning and system optimization?

That is the direction in which we are headed. However, it requires tremendous research on technologies that integrate multiple sensors, GPS, communications, control, software algorithms, real time and big data analytics, energy efficiency and management in the case of electric vehicles, etc. This would have to occur phase-wise with incremental progress that is demonstrable.  For example, Tesla, with its semi-autonomous auto-pilot feature illustrates an incremental idea that's been used by large numbers of people.  Such incremental progress is necessary for the next generation of technologies and then subsequent generations of autonomous vehicles.  We will try many approaches; some will succeed while others will fail.  We are in early stages of the beta phase of this industry.

2.      Should this network be centrally-controlled, or should it be more decentralized, with semi-autonomous sectors in close communication with neighboring and distant sectors, as appropriate?

Both.  Some decisions will be made locally - for instance collision avoidance between two vehicles requires hyper-fast response times and may even require separate sensors and circuits to respond quickly.  Other decisions, for example, travel route planning or pickup up of multiple passengers (which may be mapped to the famous traveling salesman problem or TSP) along the way by an autonomous ride sharing service may be analyzed on the Internet Cloud which may require large computing power but not instantaneous response.  ICT (Information and Communications Technologies) and IOT (Internet of Things) are key technologies to make it happen.   Or, for example, in the area of my own research on autonomous Electric Vehicles, planning the amount of energy stored within battery of an electric vehicle through the day may be a task that can be performed on the Internet Cloud whereas the power train controls in real time would be performed on vehicle - here the energy management function is distributed.  However, in the long-term as V2V and V2I communications get faster, even these paradigms may be uprooted.

3.      How can the breadth and depth of expertise and experience among the panelists be permanently harnessed in support of developing and implementing a sensible future in mobility?

My own background is in the engineering technology space having worked in the automotive sector, and also having done research in the fields of Artificial Intelligence/Expert Systems, IOT, Electric Vehicles and Smart Energy Grids, and this background allows me to contribute to autonomous electric vehicle research quite naturally. One needs multiple expertise areas to solve a problem of such a complex nature beyond simply engineering as this involves policy makers, infrastructure planners, telecommunications companies, transportation departments, technology providers, professors and students all working together to come up with innovations that can make a difference.    Technology is progressing and will progress at a very rapid pace, and so policy makers need to work with technology providers to see what's technologically possible, and how the right policies can be put into effect at the right time.  It will be an iterative and continuous process - for sure the automobile industry will experience significant and rapid change.

To watch this panel discussion, live, go here at 6:00 pm PDT on July 14, 2016.

No comments: