Although Google’s driverless car continues to be road tested with some impressive results, and research firms like Boston Consulting Group estimate that the first commercial driverless cars could appear in 2017, with more widespread adoption coming a decade later, there are still significant issues that must be worked out before driverless cars can take to the road with an acceptable level of risk.

One key is assuring that driverless vehicles can operate in a variety of climates and road conditions.

There already are cases of driverless cars encountering difficulties as they tried to navigate in the rain—or worse yet, in snowy conditions.

“We’re a lot farther from general use self-driving cars than those in Silicon Valley would like you to believe,” said Samuel Abuelsamid, a senior analyst for Navigant Research, a marketing intelligence firm based in Boulder, Colo. Abuelsamid had been driving a driverless 2015 Kia Sedona, but as he was driving in heavily snow-covered southeast Michigan, he began to notice that the inclement weather was neutralizing some of the car’s advanced technology. “The radar sensor in the front and the rear camera are completely covered,” he said. “While the snow was falling, I had to turn off the parking assist because the falling snow was triggering the ultrasonic sensors, causing the system to beep continuously while there was nothing around the vehicle.”

The adaptive cruise control in Abuelsamid’s car relies on radar, and maintains the vehicle’s speed and a safe distance from any vehicle ahead. If the car begins to draw too close to another car, the brakes are activated. Radar also detects if vehicles or pedestrians are approaching from the side, and it navigates and warns while the driver is backing out of a parking space.

The problem is that LiDAR, a sensing technology that requires light, potentially becomes ineffective and even inoperative in a heavy snowstorm. Potentially, wireless communications between vehicles and between vehicles and infrastructure road and traffic light sensors can get around this—but what do you do if wireless and/or Internet service gets interrupted?

It would seem that some kid of failover mechanism that would enable the driver to manually assume the operation of the vehicle should be part of the plan—and that significant attention should be devoted to the handling of cyber disruption or cyber security breaches that could affect software- and cyber-driven cars on the road more than rain, snow or traffic infrastructure anomalies will.

Meanwhile, Google has plans this year to put 100 of its driverless cars on the road. The cars are electricity-operated two-seaters without steering wheels, an accelerator or brakes. They only have two controls for the driver: “stop” and “go.” The technology Google employs is an assemblage of sensors, cameras and a global positioning system (GPS), with a spinning laser on the top of the vehicle that creates a 3D model of nearby objects for hazard identification. Additional front and side cameras scan for road signs and traffic lights. To assure that all of the data gets processed, the maximum velocity of the car is only 25 miles per hour—a good choice, since the mantra for now for these driverless cars seemingly is to “take it slow” and to work out the bugs.