AI-SUPPORTED ROUTE OPTIMIZATION BASED ON THE IOT SOLUTION
The general objectives of our experiment are:
- To ensure safe, cost-efficient, and reliable worldwide connectivity of the existing product of the SME, i.e.., the smart scale for measuring gas in gas cylinders.
- To support the optimization the delivery routes of gas cylinders based on the data from the smart scales for measuring gas in gas cylinders and other relevant data.
The technical objectives arising from the general objectives are:
- Integration of Narrowband Internet of Things technology into the existing product of the SME (above mentioned smart scale for measuring gas in gas cylinders).
- Development of the prototype AI-based software for optimizing delivery routes for gas distribution based on the data collected from the upgraded NB-IoT-based smart scales, customer demands, and data collected from previous deliveries.
The distribution of gas cylinders mainly relies on call centres, which means that a customer informs their gas distributor when they run out of gas. After that, a gas distributor takes care of delivering a new gas cylinder within a predefined time frame. However, this solution is rather inconvenient since gas distributors cannot plan their delivery routes, and they often make deliveries to the same places within the same day. As a result, distributors are spending too much of their resources (time, money, personnel).
The solution consists of two parts: the hardware part and the software part. The former corresponds to the NB-IoT-based smart scale for measuring gas levels in gas cylinders and sending the measured data to a remote database. The software part relates to optimizing delivery routes for gas distribution, which provides the recommended (quickest) route for the selected number of future delivery dates based on the data gathered from the NB-IoT-based smart scales and selected relevant delivery parameters. The optimized delivery route is obtained by minimizing the custom cost function that considers four main factors relevant to end-users, gas distributors, and carriers. The optimized route is visually displayed by clearly showing the starting point (i.e., a distributor’s warehouse), intermediate stops and endpoint for the carrier’s delivery day, together with the duration of the entire and each individual route.
- NB-IoT-based smart scale was tested in Slovenia, Germany and Austria in various relevant environments as well as in several operational environments in Slovenia. The testing in operational environment took place at adverse weather conditions (low temperatures, precipitation) and its purpose was to 1) test the smart scale performance in real-world environment and 2) select the most appropriate antenna type for the scale, which turned out to be the SMD antenna.
During the testing of the scale in operational environment, we tested the prototypes at diverse locations, using different gas cylinders to cover as many use case scenarios as possible. This part of the testing showed the following obstacles: the secure building had bands 8 and 20 intentionally disabled (on these two bands, the NB-IoT network in the Slovenia operates), and NB-IoT signal coverage was poor outside the city centre. The first obstacle cannot be overcome, while the second obstacle could be partially overcome by NB-IoT antenna tunning done by a professional (for the purpose of prototype development, we tunned the antennas by ourselves) and by the expected improvement of the NB-IoT signal coverage by the network provider. The measurements from the unproblematic locations were successfully fed into our developed prototype software, and we obtained the optimal delivery routes. Before the software was tested on real data, we tested it using simulated data with the purpose of setting the values of the parameters in the defined optimization cost function in a way to get the most favourable outcome.
Impact of the experiment
The experiment allowed the SME to bring their solution to a higher TRL, bringing it closer to the market stage, which is the goal of the SME. In addition, the experiment offered the SME to work with prominent researchers from the Jožef Stefan Institute, which facilitated the engagement of the SME in the field of AI. The SME and the DIH established successful cooperation and hope to cooperate with each other in upcoming future challenges.