Minnesota DOT tests autonomous shuttles in winter weather conditions
By: Sue Mulvihill, P.E., and Michael Kronzer
Automated vehicle technology is rapidly advancing and has the potential to revolutionize the transportation industry.
Significant work is being done around the country to research, test and deploy automated vehicles. However, much of this effort is taking place in warm-weather, ideal-climate states. The state of Minnesota experiences four distinct seasons, and in the winter, weather conditions can be extreme. While warm-weather climates may be ideal for automated vehicle operations, Minnesota’s cold weather offers the opportunity to truly test and advance the limits of this technology. The Minnesota DOT (MnDOT) recognized that for the state of Minnesota to realize the many benefits of automated vehicles, the challenge of winter weather needs to be tackled.
Fortunately, Minnesota has the resources and facilities to accommodate automated vehicle winter-weather testing and research. MnROAD, a pavement research facility with a 2.5-mile closed-loop track, 3.5-mile high-speed track, heated storage facilities, power, fiber-optic communications, and now perhaps most uniquely, the ability to create artificial snow and ice conditions on the roadway, proved to be an ideal testing location. With this in mind, MnDOT set out to make Minnesota one of the first states in the U.S. to reach out to the automated vehicle industry in search of a partner to test and demonstrate automated vehicle technology in winter-weather conditions.
Setting up
MnDOT started working on the project in January 2017 with support from engineering firms WSB and AECOM, and in partnership with 3M. The first step was to conduct an extensive assessment of automated bus testing and deployments taking place in the U.S. and around the world. Due to the relative youth of the technology, and MnDOT’s very recent start, it was crucial to have a base understanding of the automated vehicle (AV) industry and its capabilities before moving forward.
The AV industry research built MnDOT’s understanding of AV technology and outlined some potential automated bus and technology vendors to target as the project moved into the industry outreach phase. The vendor outreach began with an MnDOT-hosted industry webinar on project goals and preliminary requirements. One-on-one vendor meetings to gather additional feedback followed shortly thereafter. This allowed MnDOT to learn more, begin formalizing project requirements and start drafting the request for proposals that would eventually be used to procure an automated bus vendor.
MnDOT ultimately selected EasyMile (partnered with First Transit). EasyMile provided a level 4 automated shuttle, the EZ10, for winter condition testing and public demonstrations from November 2017 to April 2018. The “level” for an automated vehicle refers to the levels of automation set by the Society of Automotive Engineers (SAE). Level 4 means the vehicle is highly automated and relies solely on the technology to perform the driving task with no human intervention. Although it can operate without a human driver, it must use predetermined routes and operate in environmental conditions that are deemed safe. MnDOT determined that testing in winter conditions will be increasingly important in order for AVs to be designed to perform safely in all types of conditions.
Testing it out
Winter testing began almost immediately at MnROAD. In November 2017, winter arrived early and provided the much-anticipated natural snowy conditions needed for testing. The automated shuttle ran through various scenarios that a vehicle would typically encounter daily on a public roadway in normal operations. The goal was to observe and document the vehicle’s behavior in ideal weather conditions, and compare these findings to the vehicle’s behavior in poor weather. The scenarios included vehicle starting, stopping, following its programmed route at varying speeds, and reacting to conflicts such as work-zone barrels, pedestrians, bicyclists and passenger vehicles. The tests were then replicated in different winter conditions (snow falling, snow on the pavement, windy/blowing snow, ice on the pavement, extreme cold and salt spray on the vehicle). Observations also were made during different times of the day (morning, mid-day, evening and night) to see potential impacts from different lighting conditions.
MnDOT’s project team also realized that Minnesota winters are unpredictable, and natural snow and ice cannot be guaranteed. There was a risk that winter condition testing would not occur unless winter conditions also could be artificially created. MnDOT and its consultant project team reached out to a local ski hill to contract snowmaking services. In January 2018, two separate snowmaking machines were positioned along MnROAD’s test track and produced artificial snow to continue winter condition testing. While artificial snow differs in texture from natural snow, the challenges to the automated vehicle remain largely the same. The snowmaking equipment allowed the project team to create controlled falling/blowing snow conditions and to create snow on the pavement at varying depths.
The shuttle was demonstrated live in downtown Minneapolis in the days leading up to Super Bowl LII. Public engagement was high and response was largely positive.
Going public
The second major project phase involved demonstrating the technology to the public in downtown Minneapolis in the days leading up to Super Bowl LII. Public outreach and education surrounding automated vehicles is key to the advancement of this technology. The anticipated safety and mobility benefits of AV technology require strong public understanding and acceptance, and already it has been a tough sell. Super Bowl LII taking place in downtown Minneapolis provided a fantastic opportunity to showcase the automated shuttle and to let the public experience the technology firsthand.
The public demonstrations took place on a closed street in downtown Minneapolis the weekend before Super Bowl Sunday. The public interest was tremendous. The shuttle provided rides to nearly 1,300 people over three days and garnered significant local media interest. MnDOT also gathered feedback from the public through a brief mobile application survey. The survey involved simple questions to gauge public perception of automated technology, such as “Having ridden the driverless vehicle, do you think the ride was safe?” (96.4% of respondents answered Yes); and “Are you looking forward to having driverless vehicles operate on all roadways in the future?” (83.6% of respondents answered Yes).
Overall response was very positive and supported anecdotally by conversations the project team had with the public during the demonstrations. MnDOT’s biggest realization was that public outreach and education are necessary to drive acceptance of the technology. It also was a reminder of how powerful it can be to let the public engage with AV technology directly.
MnDOT’s automated shuttle pilot project was a learning experience for all involved. The big takeaway was that the EZ10 automated shuttle performs well in varied pavement conditions such as wet or snow-covered pavement, but still faces some challenges when encountering falling precipitation such as rain and falling or blowing snow. The vehicle’s sensors are extremely sensitive and pick up their surroundings with great accuracy. This is great from a safety perspective, and the vehicle itself is programmed to be very cautious. The challenge that snow creates is that the sensors detect each snowflake that passes in front of the vehicle. The vehicle then must interpret and decide if what the vehicle sensors “see” is a hazard it needs to slow down for or something it can ignore. When in doubt, the vehicle always takes the conservative and safe approach, often slowing down significantly in snow. Shuttle engineers collected sensor data throughout the falling snow tests and are refining system algorithms to make the vehicle better at filtering out objects that are not hazards (e.g., falling snow) while still reacting appropriately to objects that are hazards, such as pedestrians, vehicles and bicycles. Options to further harden vehicle sensor housings for winter weather conditions also were discussed for future AV shuttle models.
This project is a major first step, in what will prove to be a long journey towards tackling the challenge of operating automated vehicles in winter weather. It is a journey that will require collaboration across the industry from both the private and public sectors. Minnesota looks forward to identifying more operational tests that span its seasonal weather challenges as residents prepare for the use of automated vehicles across the state.
About The Author: Mulvihill is deputy commissioner and chief engineer, and Kronzer is CAV project manager, both for the Minnesota DOT.
