Toyota wants a co-pilot to take charge of your car
From the starting of the year, we are constantly seeing the news about autonomous vehicles and how they are going to dominate the market in the future for better and safe driving. Ford, Google or even Apple is flashing into the technology news about their versions of autonomous cars and companies like NVidia and Qualcomm are already working on custom chips to leverage the idea of a driverless car. It might not sound surprising that Toyota is also jumping into the pit of driverless car battle and is in works of developing a feature called “guardian angel” that is going to wrest the steering wheel when it senses some danger.
Sounds like that this co-pilot thing is a level lower than the fully autonomous cars but the Japanese carmaker sees the combination of human and machine driving as a key step towards complete autonomy.
The simple concept is to think of a virtual driver or co-pilot that is sitting to take control temporarily whenever its wants to avert danger that is beyond the scope of human abilities, explains the CEO, Gill Pratt, of Toyota Research Institute which was created last year with a $1 billion in funding to propel the research in robotics, artificial intelligence and automated driving. The working of this system is analogous to the anti-lock braking and emergency braking systems.
Understanding and managing the driver behavior might be the new challenges faced by the company in their approach. The researchers plan to test the technology near Mt. Fuji in Japan in a giant moving simulator. The setup simulates driver realistic scenes which allows to record how drivers react in such situations and how the car should overtake the steering wheel because it knows better when avoiding accidents and does not know how to panic. This makes it a car with partial autonomy just like power steering and lane-departure prevention.
The testing shows that the drivers take around 8 seconds to regain their focus which makes the car to revert to the manual control within 8 seconds which can be achieved in gradual improvements. To prevent the usage of immense power as the computers in the driverless cars use, Pratt also highlighted that the technology could use neuromorphic chips, as conventional computers do.
You might not know it but Toyota is working on self-driving cars for over a decade but it still lags behind when it comes to testing in real world scenario. However Pratt says that the data that efficiently trains the algorithms makes self-driving possible can be taken from the equally real virtual environments. He says that the Toyota cars will need to be tested for real world application until they cover a trillion miles which can be achieved through virtual environments.
Pitt also says that for a truly intelligent autonomous cars, the ideal scenario can be the companies working in the self-driving domain come together to share their data – both obtained from real and virtual environments – so that the others can also learn from it.