A project I had the pleasure of working on back in 2015 has won the Alberta Transportation Minister’s 2018 Award of Excellence for Safety Innovation.
The system is known as the Road Condition Warning System (RCWS) and was installed as part of the larger Road Weather Information Stations (RWIS) expansion project for Alberta Transportation. It had a broad objective: “To install a warning system that used the current wind speed to warn drivers of the danger, and to encourage them to take alternate routes”.
A solution was needed for Alberta Highway #22, between kilometers 7 and 27, which is known to the locals as the “Wind Tunnel”. Thanks to local geography and Chinook winds, strong winds gusts were blowing vehicles over or into on-coming traffic in the other lane.
Between November and April, motorists on Hwy 22 are at highest risk. On just one day, February 11 2011, approximately eight vehicles were blown over, including 6 semi trailer units, 1 pick-up with a utility trailer and 1 R.V. unit.
Between September 2010 to April 2011 there was 16 similar accidents, and previous years had 4-6 each, which each accident costing from $25k to $40k, and an extreme risk to life and limb.
My involvement on the project started at the very beginning: the proposal. From there, I held many roles: project lead, acting project manager, lead engineer, software development, vendor relations and procurement. The implementation of the RWIS expansion system was to be over several years, and so the first thing we did for the Warning System project was to upgrade the wind sensor at the Lundbreck RWIS site from the standard requirements to the higher-end, all metal construction Lufft VENTUS sensor, which is designed to not suffer from icing. The warning system would rely entirely on the wind speed measurements from the RWIS site, so it had to be reliable.
It was a year or two later, after most of the RWIS expansion sites had been installed that we returned to the warning system. Since the requirements were so broad, we met with the various stakeholders to determine the actual requirements. This was the most time-consuming part of the project, but also the most rewarding as it exposed me to new lines of thinking and put me in touch with other experts in different disciplines. Stakeholders from my company DTN, Epcor Technologies, Alberta Transportation, the Lethbridge district, and IBI worked together to define the problem, outline the solution, and tailor it to the situation.
Two areas that were probably the most interesting to me were sign design, and algorithm design.
The process of designing the sign was particularly fascinating. We all drive past road signs all the time, with very little thought about how they are designed. However, you can’t noticing a badly designed sign!
Alberta Transportation’s road safety department has guidelines regarding font size, vocabulary and graphic design, which we had to take into account. These standards ensure that a passing motorist travelling at 80 km/h would be able to completely understand what the sign was telling them, without ambiguity.
We decided on two sign design types: one with an active speed display, and with without. The active speed display was a 3-character Dynamic Message Sign embedded in a large static sign. Both signs had amber warning lights on the top, which would flash in an alternating, “wig-wag” pattern.
Thanks to the design of the back-end system, we were able to support these two different sign technologies with comparative ease.
For example, the solar powered sites did not have an NTCIP interface, instead we designed simple Digital Output circuitry that controlled relays and was interfaced to the modem, which in turn was controllable with Modbus – a standard protocol in the SCADA industry available to us on the OASyS backend.
This resulted in significantly reducing the budget for the signs, since NTCIP compliant signs are quite expensive, and far to complicated for such a simple design. It also allowed future sign technologies to evolve and be used to expand the system. This in fact has already happened, with 2 bridge deck warning systems being deployed in Dunvengan and Drayton Valley, which combine pavement sensor data with visibility data to evaluate a risk index, and warn drivers using a 3-line, full-text display on both sides of the bridge approach.
One of the characteristics of this system that differentiated it from others like it was the distance between the various components of the system.
The RWIS station was located in the trouble spot, where the wind speed was judged to be the most dangerous. In a study conducted by IBI, it was found to be the location that correlated with the highest occurrence of weather related accidents.
The signs are located at strategic turn-off points, giving motorists the opportunity to take alternate routes, to avoid the danger.
In all, 6 signs were installed over a 100km stretch of Highways 22, 533, and 3. They were all connected via cellular modems to the back-end system, with high gain antennas and signal boosters to ensure a strong steady signal in an area of Alberta were frequent drop-outs are common.
Thanks to the collaborative nature of the project, we were able to incorporate an important feature brought forward by the Highway Maintenance Contractor: local control and override. In this way, the person in the field can have ultimate control of what the sign should display, which is an important safety consideration, since motorists will quickly ignore signs that are repeatedly wrong.
One of the most important design requirements we needed to get just right was the way in which the system interpreted the real-time wind-speeds, and how it would use them to control the sign.
By its very nature, wind is highly variable, and even more so in this region along Highway 22, where the wind gusts can spike from 20 km/h to 90 km/h in a matter of minutes.
The concern was that if the system was too sensitive to the wind speed, it might turn off the warning signs just before a motorist drove by, giving them the impression that nothing was wrong, only to have the wind speed pick up again after it is too late to activate the signs. On the other hand, signs that turn on and off very rapidly will reduce a motorist’s trust in the system, reducing its ability to influence their decision.
We explored various control schemes, data dampening methods, timing thresholds, sign designs, seasonal variability and driver cognition before settling on the final system.
In the process, a simulator was developed that was used to demonstrate the various control scenarios using real wind speed data. This proved very valuable as a tool to visualise the system’s outcomes, instead of interpreting the raw numerical algorithms.
The simulator was an excuse to learn some reactive web programming, and was a blast to put together. As the user moves the mouse over the plot, they simulate time passing, as witnessed by the warning lights on the sign turning on and off as various thresholds are adjusted. Adjusting the thresholds allowed the user to immediately see the result on the sign.
We found that the stakeholders were immediately comfortable with this visual approach, and found it much easier to make a confident, informed decisions about the final requirements for the algorithms, particularly because they often weren’t familiar with terms such as “dampening”, “chatter”, “asynchronicity”, “hold timers”, etc.
The result was that many of the potential control and configuration strategies were eliminated as being unnecessary or too complicated, and a single threshold was used instead. However, the system remained ready to handle the more complicated scenarios where surface sensors and visibility sensors were combined at the bridge deck sight in Dunvengan and Drayton Valley.
In the end, the overall system operation is quite simple:
Of course, the details were quite complex.
- Real-time data from field devices
- Configuration smoothing and hold-off periods
- Manual tiggers and periodic events
- Generic evaluators on inputs
- Commands to field devices
- Triggers of other actions, to create chains of actions called sequences
- Rules being to a priority group, and have a unique priority within the group
- The highest priority rule in a group winds, and only its action is executed.
The benefits of the rule engine that it allowed the customer to continue to experiment with new algorithms for the automated system. New rules can be added and existing rules modified without any downtime, and without any code being altered.
It also meant that inputs from other sources could eventually be taken into account. For example, the thresholds could be modified based on a forecast. Or, the signs could be switched on if Environment Canada issues a high wind warning for the area.
It also meant that we didn’t have to send programmers to the site, and have them troubleshoot the algorithm when peering into a laptop screen in the full sun while standing in the ditch.
The system has reduced rollovers by 90 per cent in the Lundbreck area, and according to the local Alberta Transportation district staff and their Highway Maintenance Contractors, it has dramatically improved the safety of the road.
|High Wind Events||85|
|Number of Hours the Beacons were active||164|
|Number of high wind- related incidents or crashes (down from 35-40 previous year)||4|
|Highest Wind Gust Detected|
(*)Updated Oct 25, 2019
Pictures of these signs regularly appear on Social Media, as locals use them to record the wind gusts in their area, and alert fellow users that they should perhaps find an alternate route for that day.
I am very proud to have been involved in this project.