The industry may be abuzz about the promise of self-driving cars, but we've yet to see them tackle some of the harshest weather conditions. Instead, like winter birds, autonomous vehicles prefer warm, dry weather.

That's not the case in Singapore, which is home to the Center for Excellence for Testing and Research of AVs, or CETRAN. IEEE Spectrum last month received an in-depth look at how the facility tests self-driving cars for one type of severe weather: the monsoon. The tropical storms are common in the country and often lead to flash-floods.

CETRAN is equipped to mimic these conditions perfectly. Aside from recreating everything autonomous cars find in the real world, its facility has a mammoth shower head of sorts that sits high above the test center's roads. The metal frame has the ability to pump 9 inches of rain per hour for testing purposes.

Self-driving car companies, such as the startup nuTonomy, use the facility to test their technology.

As it stands, self-driving car radar, sensors, and even lidar are not capable of handling monsoon conditions or even heavy rain. The rain drops can absorb radar waves and laser beams from radar and lidar systems, which causes attenuation, or the droplets can reflect these signals and register as obstacles. Rain can also register as fog on the cameras used by self-driving cars. Ensuring an autonomous car can detect standing water is enough of a challenge, let alone making the car smart enough to know how deep the water is.

Yet, progress has already been made at the test center. Using machine learning, developers chose fixed points in the test center such as tall buildings, poles, and tree trunks. Even in poor weather, the large objects still provide the vehicle with data. Another team engineered a de-hazing filter that allows cameras to work better.

There's obviously a lot of work to be done with monsoons and rain in general.

That will leave another major issue for self-driving cars to overcome in the real world: snow storms. CETRAN wants to tackle that as well by working with other autonomous vehicle test facilities, such as those in Michigan and Sweden.