OSS.5 USA is America’s first platform bringing together all stakeholders who play an active role in the achievement of high-level functional, system & operational safety for fully automated driving. In the run-up to the event, we.CONECT spoke with Avinash Balachandran, Senior Manager at Toyota Research Institute about new technical innovations, latest updates to standards, and pressing challenges regarding operational & functional safe systems in Level 4 & 5 vehicles. we.CONECT: What are your main responsibilities in your current role? Avinash Balachandran: I’m a Senior Manager and lead the Extreme Vehicle Dynamics Control team. My main responsibilities are leading hardware, software and operations teams to investigate advanced topics in motion planning and control. This includes exploring motion planning and control in extreme environments, incorporating learning into traditional approaches and sharing control with the driver. I also work closely with Stanford University where I’m a co-PI with 5 other faculty members on TRI-sponsored research projects. Through these efforts and working with other Toyota entities, I bring innovative ideas to production and make cars safer. we.CONECT: What technical challenge fascinates you most about autonomous driving? Avinash Balachandran: Automated driving has so many challenges so it’s hard to choose. One challenge that is part of what I study is how best to incorporate expert driver data into our algorithms to improve performance. In particular, the best human drivers (like race drivers, chauffeurs etc.) are really good at reading the situation, making good decisions that balance safety and vehicle performance as well as executing precise and robust controls especially in traction-limited events like driving on snow. Learning how to utilize data from expert drivers and encode their techniques, experiences and control into automated driving is a challenge and one which we believe can make vehicles safer both in Chauffeur (fully self-driving) and Guardian (ADAS) applications. we.CONECT: Where do you see the biggest opportunities in the application of AI, machine-, deep- & reinforcement learning in the development of fully autonomous vehicles? Avinash Balachandran: The opportunities we see to use more data-driven techniques are in using expert driver data to we.CONECT: How effective are virtual testing, simulation and modelling in Level 4&5 today? What should be improved? Avinash Balachandran: Virtual testing, simulation and modelling are a critical part of autonomous vehicle development, testing and verification. These technologies allow us to multiply our effort beyond physical testing to ensure we have the best and most robust coverage over the problem domain. The main challenge here is getting the simulation environment to better reflect reality. This includes improving the physical models to ensure that the vehicle obeys the same physical laws (especially at the limits of handling) to improving motion models for other road users as well so that the traffic scenarios more accurately reflect reality. we.CONECT: What have been the highlights of your work with AI/machine learning in the development of autonomous vehicles so far in your career? Avinash Balachandran: There are a few things that I’m particularly proud of. Firstly, my team helped develop the core shared/blended control algorithm (Envelope control) that powers Toyota’s Guardian technology (see CES 2019). Being able to share control with the human driver allows for the melding of the strengths of both humans and AI to enhance safety. More recently, my team has developed core motion planning algorithms that incorporate expert driver skill (see article) into the vehicle. This allows vehicles to have the same reflexes and insights as expert drivers and can improve vehicle safety and performance. Imagine a world where an expert driver has your back when you’re in a tough driving situation! we.CONECT: How is your company developing its AI/machine learning/deep learning capabilities? What are the challenges? Avinash Balachandran: One of TRI’s core goals is to transform Toyota into a more data-driven company. As such, we look at incorporating AI/ML in various forms. In particular, TRI works on Automated Driving (both Chauffeur and Guardian), Robotics, Advanced Materials Discovery and Machine Aided Cognition. They are all built on the idea of leveraging AI/ML as best as we can. The challenges to this are many fold. Some of the big ones are obtaining the right kind of data for tasks at hand, making these approaches robust/reliable for safety-critical applications and making these we.CONECT: What would you change in the AV industry to speed up and improve autonomous R&D here? Avinash Balachandran: There are many challenges in developing AVs and having each company solve all of them separately is difficult. We think a useful change would be for the various companies in this space to identify ways to collaborate on certain topics like safety and validation so that we can tackle some of these big challenges together and pool our resources. we.CONECT: Please explain in brief the key aspects of your session at the OSS.5 USA 2021. Avinash Balachandran: I will be speaking on “Learning from Expert Drivers to Improve Vehicle Performance and Safety”. The best human driver’s have skills and abilities that enable them to control the vehicle at the edge of performance. Learning from them and harnessing this skill allows our increasingly automated vehicle systems to perform better in challenging conditions like driving on snow/off-road and performing aggressive emergency obstacle avoid maneuvers. In this talk, we will look at what we can learn from expert human drivers. We will also show techniques developed at TRI which allow us to better harness the skills of these drivers to improve vehicle safety. we.CONECT: What expectations do you have towards the OSS.5 USA 2021? Avinash Balachandran: I am excited and honored to participate in OSS.5 USA 2021. I look forward to learning best practices from my colleagues and identifying opportunities for collaboration. we.CONECT: Thank you for the interview! To all our readers, if you are interested and want to join us together with Toyota Research Institute and want to learn, engage and discuss automotive tech innovation in real-time with thought leaders across the globe, feel free to save your seat:https://www.oss-5-usa.com/book-now
augment our current approaches in perception, prediction and motion planning. In particular this
data can help us build algorithms that can predict the intent of other road users better, make
better decisions and create safer/more naturalistic driving styles.
approaches interpretable so they become less of a blackbox.
