Road dirt, snow and ice could clobber car sensors


Road dirt.

It is hard to imagine with all the research into connected and self-driving vehicles with multiple sensors and wireless connections that road dirt would be so important.

But it is-- along with road tar, bugs, rocks and ice and snow.  All the various sensors envisioned for future self-driving cars could be occluded or damaged by dirt or rocks, making their perceptions of obstacles or conditions unreliable. 

A damaged sensor could in turn lead to a false reading, which might cause an accident. It’s a serious concern and one that Roger Melen, a senior advisor at Toyota InfoTech Labs, is researching.

Melen has a long history as a researcher, having applied for 60 patents over his 44-year career. He first studied CCD image sensors when doing PhD research at Stanford University in the 1970s. He also taught the electrical engineering entrepreneurship class at Stanford from 1976 until 2019.

“One long-term concern for reliable and safe operation of systems in cars is that these radar and camera sensor devices have unobstructed observation of the roadway in various directions,” he said.

“Currently, cars are cleaned and we are not yet highly dependent on these systems. As times goes on, there will be more need that onboard sensors operate in an unobstructed view,” he added.

The answer to the obstruction  problem will likely be fast, low-latency wireless connections to other vehicles and to roadside sensors that transmit data about pedestrians and road conditions.  By hearing from other nearby vehicles traveling in the same or different directions, a vehicle will be able to compare the readings from its own sensors to the readings from sensors on nearby cars.

“At the very least, we want onboard systems to be aware that these sensors are having a current problem which may alter the performance of the system,” Melen said.

Many high-end cars today already include multiple sensors to aid Advanced Driver Assistance Systems (ADAS) such as lane correction or obstruction detection.  Some current and future models will have 10 or more sensors int he form of cameras and radar pointing in all directions.

At Toyota, a Digital Communications Module has been standard in North America models in recent years to allow talk between vehicles.  Denso provides the module with cellular, Bluetooth and Wi-Fi connectivity. Newer cars will have an even more advanced form of vehicle-to-vehicle (V2V) and vehicle-to-everything (V2X) communication “to make it easier to talk between cars,” Melen said.

“Certainly through the cloud today we can obtain information on cars that are near us that recently should have received the same sensor environment and we should tell to some degree if our car or their car is having impaired conditions such as bugs on the windshield reducing performance,” Melen said.

Higher speed and accuracy in the future will help, since sensors can be used to automatically reduce the speed of a car if it is about to rear-end the car ahead. Those sensors must operate in a high degree of performance and not suffer from impairments such as a coating of mud or ice.

With rapid V2X communications “we can develop software and apps to talk between cars and share algorithmic information about what our sensors see,” Melen said. “Higher performing verifications can be built if good vehicle-to-vehicle and vehicle-to-roadside information is available.”

“Advanced sensors in cars are likely to be increasingly important in the future, but overall performance of these systems is highly dependent on how well these sensors perform,” Melen added. “Building V2V communications is a very important innovation that will ensure good ability to validate the performance of onboard vehicle sensors.”

Roger Melen, senior advisor at Toyota InfoTech Labs, will appear on virtual panel Monday, December. 14, as part of AutonomousTech Innovation Week. The panel kicks off at 11:45 a.m. EST on Monday on the topic of "Innovations in Autonomous Technology: Sensors and Beyond."  The overall event starts at 11 a.m. EST on Monday with keynote addresses. Registration is free and online.