In the first part of this series, we talked about our search for innovations aimed to reduce fuel consumption and its cost. In this article we will take a look at a wider range of ideas and projects related to our truck fleet — from electric vehicles to fatigue monitoring wristbands. Of course, as in the first part, we will be sharing details on specific projects and corresponding results.
As for participants — they are still the same: our transport department, together with the innovation team. So, let’s get started!
Repairs can be predictable, right?
Or, on the contrary, they can’t? As you might know, trucks need to be repaired not only because of accidents, but also because of the natural wear and tear of their components and assemblies. Therefore, at some point we decided to learn whether it was possible to predict, based on some direct or indirect indicators, that a particular component or assembly unit would soon fail, and to prevent its breakdown.
We made an assumption that predictive maintenance systems could help us with this. Such systems perform remote vehicle diagnostics in real time (checking the status of various systems and identifying potential faults), and also create individual maintenance plans.
We tried a couple of predictive maintenance solutions from well-known foreign vendors, but they did not pass our proof-of-concept stage: they analyzed data from our telematics system for different truck models, but it turned out that the volume and speed of the data received did not allow these systems to make reliable predictions. To improve the outcome, they required widespread installation of very expensive telematics units, so we decided not to continue with the deployment of these systems.
Simultaneously, we decided to test several domestic systems of this class and even launched a pilot: we received data regarding various technical faults via the CAN bus, processed these data and compared the remaining lifetime of the car units, and then compared the obtained results with actual faults. In the end, we came to the conclusion that the current level of development of such predictive maintenance technology does not allow us to project breakdowns with sufficient accuracy. However, we decided not to give up and continue to search and test similar solutions.
Electric trolley is the king
You’ve probably seen hydraulic trolleys more than once — they are used in stores for transporting pallets. It goes without saying, we also use them everywhere. But one day we decided to take a step forward and try electric trolleys instead of the old-fashioned hydraulic ones.
It was the night shift unloading that served as the catalyst for this project. The thing is at night drivers have to not only deliver the goods to the stores, but also unload the cargo themselves (since there is no one in the store to help him).
We assumed that electric trolleys should be able to solve three problems simultaneously: they would reduce the burden on the drivers, speed up the unloading process and allow drivers to deliver more goods within one haul cycle.
The thing is a hydraulic trolley can carry a pallet of a smaller volume, with a weight limit of 500 kg. Electric trolleys, on the contrary, can accommodate larger pallets, and the permissible weight is almost twice as large — 900 kg.
The pilot test proved our hypothesis true; everyone liked the equipment. As a result, we run a tender and introduced electric trolleys into the purchasing standard for our Chizhik and Pyaterochka stores.
But sometimes the economy doesn’t add up
A few words about theft prevention. Our search for ideas and innovation in this area led us to electronic locks that could potentially replace the conventional lead seals used on our trucks.
Their key advantage is the ability to program these locks in such a way that they open only in the area of loading and unloading of goods. Accordingly, this would prevent drivers and any other third parties from accessing our goods.
However, when working on this concept, it turned out that it was not worth it, so we refused to continue with this project.
Besides repairing and refueling our cars, we also need to ensure that our drivers treat vehicles with proper care; we should help drivers avoid fatigue while driving, and ensure they don’t get into accidents that for one reason or another depend on them.
This being said, fatigue control is one of the keys factors of accident-free driving, so it is important to constantly monitor the condition of drivers and prevent excessive fatigue. In our search for innovation in this area, we’ve tried five different solutions, and in the end, we are implementing just one of them.
Driver’s bracelet
One project that never passed the proof-of-concept stage involved special wristbands for drivers to monitor changes in their skin conductivity. As you might know, this indicator strongly correlates with the person’s tiredness and can be used to detect fatigue.
Our system was based on the following algorithm: the bracelet analyzes electrodermal activity (EDA) and if it falls (which is a sign of falling asleep), the bracelet would begin to vibrate, preventing the driver from falling asleep. In addition to vibration, a special siren signal also comes from the speaker in the cabin.
However, after testing this system on three cars, we came to the conclusion that this technology is not quite efficient: the siren is too quiet, and the vibration on the wrist is sometimes hard to feel while driving. In addition, it is impossible to control if the bracelet itself is still on the driver’s wrist, not to mention difficulties with passing on the bracelets between drivers.
To make things even more complicated, let’s add some cameras
Since it was not possible to solve the problem without much pain (so to say), we began to look towards more complex systems that involve installation of various cameras and sensors. Their task is to recognize and prevent not only fatigue, but also a number of other potentially dangerous actions (or inactions) of the driver.
Among them, for example, such cameras and sensors are able to detect an unfastened seat belt, an unsafe distance from the car in front, changing lanes without a turn signal, sudden maneuvers, talking on the phone (or moments, when drivers got stuck in their phones), eating while driving, and so on.
In addition, such a system should monitor whether the driver has blocked the camera (on purpose or accidentally), and, of course, monitor his own physical condition — detecting prolonged head turns, eye-blinks, yawning and head nodding.
We’ve carried out a series of experiments with four similar systems. But only one of them has proven effective and was able to properly detect drowsiness and fatigue. The rest of them either had false positives (reacted to drowsiness where there was none, or, on the contrary, were unable to detect fatigue, when drivers were actually tired), or were not effective enough in comparison with the ‘winner’.
In total, the pilot project featuring three systems (the fourth one was rejected at the PoC stage) covered more than 100 trucks. The solution we implemented in the end showed a reduction in accident rates by more than 2 times and a reduction in financial losses by almost 98%.
A nice bonus of the implemented system is a rear-view camera, which helps drivers maneuver in residential areas and near retail outlets. Moreover, constant video surveillance of the car helps prevent thefts.
At the moment, this project is being scaled up, and we are installing this system on all our commercial vehicles. Besides, it has also been introduced into our standard for procurement of new vehicles.
As we said at the very beginning, any failed experiment is, first of all, experience and lessons learned, which among all help us come to successful implementations. And any successful one is a step forward, giving us further motivation to run even more tests, implement even more projects and generate even more ideas.
