Hampson, Michelle. “A New Way For Autonomous Racing Cars to Control Sideslip.” IEEE Spectrum: Technology, Engineering, and Science News, 21 Sept. 2020, 9:00 GMT, spectrum.ieee.org/cars-that-think/transportation/self-driving/get-a-grip-a-new-way-for-autonomous-racing-cars-to-avoid-sideslip.
Article title:
A New Way For Autonomous Racing Cars to Control Sideslip
Photo: Vivek Maru/Formula Student Germany
My summary of the article:
Race car drivers rely on their gut feelings based on their experience to make tight turns during high speed racing possible. Of course, this is not possible for autonomous racing cars due to the absence of a driver. Consequently, current autonomous racing cars rely on special sensors that are extremely accurate yet highly expensive and heavy to make tight turns while minimizing loss of velocity. This approach is very expensive and adds extra load to the vehicle; this motivated the Autonomous Systems Lab at ETH Zurich to create a new design with rather simple sensors that does the same job with less money and extra weight.
Basically, the new design operates by training the vehicle's motors' reactions to multiple measurements taken by simple sensors. The training is based on empirical data from actual racing cars on surfaces with different properties such as flat, rough, bumpy, and wet. In other words, it implements machine learning, a basic form of artificial intelligence, to the operation of autonomous racing cars.
This system, however, is not yet perfect; Sirish Srinivasan, a member of the development team - AMZ Racing - at ETH Zurich, suggests that implementing machine learning may be a double-edged sword since although the approach works well in trained circumstances, it may severely lack the ability to react to untrained situations such as unusual weather conditions, changes in tire pressure, and more. Consequently, the team believes that making the autonomous system able to react to unexpected circumstances is the biggest challenge they must overcome.
My response to the article:
The idea of autonomous racing cars is indeed fascinating. I have watched Formula 1 racing since I was young, but to imagine them being done without human drivers is of greater interest to me. What's more intriguing about this is that researchers have found a way to implement artificial intelligence in the form of machine learning in developing this model in order to eliminate the high cost and extra weight (load) of literally considering almost every variables that may be taken account to calculate the intricate operation of the motors. Although the researchers mention that this method is not yet perfect, with more research, experimentation, and time, I believe that what researchers learn during this may lead to the evolution of autonomous driving systems - not only in racing but also in automobile driving for civilians.
Comments