Paper Title (use style: paper title)

1 downloads 0 Views 218KB Size Report
In this work, a smart winter service platform is designed for the transparent ... Winter service operations are classified as chemical, ... proposed an ARM-based hardware for vehicle monitoring, ... constructive, composite and metaheuristics categories, and ... Kafka message brokers on different topics which are used by.
A Smart Winter Service Platform and Route Planning Algorithm Perin Unal, Yunuscan Kocak, Yunus Donmez Teknopar Industrial Automation Ankara, Turkey {punal,kocak,donmez}@teknopar.com.tr Abstract—Winter service operations are one of many responsibilities of local authorities essential for public safety. Consisting of snow removal and spreading de-icing material, operations require the assignment of significant resources in terms of vehicles and personnel. For that reason, the development of tools enabling authorities to engage in the transparent monitoring and planning of the snow removal and de-icing operations is beneficial for the community. In this work, a smart winter service platform is designed for the transparent monitoring of the operation and a vehicle route planning algorithm is designed as a decision support system for the operation center. The platform consists of an electronic control unit attached to the vehicle for monitoring and data collection, which is transferred to the cloud for analytics and reporting purposes. In the cloud, distributed big data technologies are used for fault-tolerance and fast processing. Web and mobile applications have been developed for visualization and increasing public awareness of the operations. Keywords-geographic information system; internet of things; cloud computing, smart environment

I. INTRODUCTION In order to ensure that citizens are safe outside in winter, many countries make an effort to provide winter maintenance operations. In particular, snow and ice removal are essential for the safe functioning of roads and continuation of everyday life. Ice removal operations consist of spreading chloride salts, de-icing solutions and plowing snow. For these purposes, local authorities have specialized vehicles which can spread salt/solution and plow snow at the same time. The direct and indirect cost factors must be taken into account when planning winter service operations. The direct costs are the amount of de-icing solution, salt, fuel and the number personnel required for the operation. There are also fixed costs related to the vehicles, depots, and related equipment. In terms of the indirect costs, salt has a harmful effect on other vehicles and road infrastructures, and in addition, it has a negative impact on the environment since it mixes with underground water sources; therefore, it is important that salts are used in appropriate amounts. Currently, data collected from winter service vehicles are limited to the total fuel and material usage per vehicle, which

makes it hard to assess other cost factors in relation to the condition of the vehicle, routes to be covered, and the weather conditions within a given timeframe. To reduce costs and improve performance, there is a need to measure the efficiency of winter service operations in a bottom-up approach, starting from efficiency of the vehicles to the planning of the operation. Recent developments in the Internet of Things (IoT) enable the creation of smart vehicles, which can transmit real-time data to operation centers for further analysis. This allows operation centers to have a transparent view of the current status of the vehicle and detailed performance report of the vehicles and personnel, then compare this data with those of other vehicles to gain awareness of problems and use it for planning the next operation. In planning of the routes, the operation center uses existing knowledge and experience without knowing the optimal solution. There are also other factors besides costs that need to be considered in planning, such as the importance of the roads and the speed of the operation. Often, too many resources are assigned to one task with too few resources remaining for other tasks. Therefore, route planning algorithms should be able to use data taken from vehicles for the optimization of the operation. Another important aspect of the current winter maintenance operations is that citizens cannot access information concerning the status of the snow clearing and de-icing operation. Therefore, offering accurate up-to-date information about the cleared roads, exact position and route of winter service vehicles will help citizens to make informed decisions concerning their choice of transportation route. This would increase safety and travel efficiency. In this work, a Smart Winter Service Platform (SWSP) is presented. The proposed SWSP consists of an electronic control unit that is attached to the vehicle and transmits sensor data into the cloud, a distributed big data processing and storing platform, a smart routing algorithm for vehicle route optimization, and web and mobile applications. The remainder of this paper is organized as follows: In section 2, related work is presented, particularly related to winter service operations, vehicle monitoring systems, distributed architecture, and route planning. Details of the platform are discussed in section 3, the smart routing

algorithm is presented in section 4, and section 5 contains the conclusions. II.

BACKGROUND & RELATED WORK

Winter service operations are classified as chemical, mechanical and thermal depending on the method of operation [1]. In chemical operations, a chemical is spread onto a surface to melt existing ice and prevent future formation of ice. Salt is most commonly used for this process since it is easily available, low cost, and effective. Mechanical operations consist of plowing and manually removing snow as well as hauling away for disposal. In thermal operations, heat is applied to remove snow and prevent icing; however, this method is costly, and thus it is not widely used. In chemical and mechanical operations winter service vehicles and personnel can work together. In addition, winter service vehicles are equipped with material storage and mechanical tools for plowing operations. The operator has a control unit in which amount of salt spread and position of the blades can be organized. In operations, the operator is given a route to clear and returns back to depot. Monitoring of the vehicle is beneficial in understanding the environment as well as the vehicle status. In vehicle monitoring systems (IVMS) facilitate the use of geographical information systems with an attached electronic device that can track and monitor vehicle status. The status can include coordinates of the vehicle, fuel, and oil condition and data taken from its environment. An IVMS can be active or passive depending on the type of data storage and communication. Data is usually sent to servers for further analysis via a cellular network if available or stored in the device for uploading later. Pearn and Wu [2] proposed an ARM-based hardware for vehicle monitoring, which is equipped with sensors, GSM and GPS modules. Data is stored in a memory card and can be accessed at any time. A platform proposed in [3] was composed of communication, application, and database servers using client-server communication on TCP/IP and UDP. The real time monitoring of many vehicles requires significant computing power; therefore, a distributed architecture is ideal for real world problems. Distributed technologies have two main advantages; first, they are fail-safe in terms of unexpected breakdowns since data is replicated among instances, which reduces the possibility of losing operational data. Second, distributed architecture in processing means accessing more computing power; therefore, calculations are faster. Distributed messaging systems have been developed for efficient messaging, and they are used as a middleman among message producers and consumers. In Apache Kafka [4], a producer can publish messages concerning a topic that can be accessed and utilized by subscribers. Load balancing is implemented by dividing a topic into multiple partitions where each broker stores one or more parts. Distributed data processing consists of processing batch and streaming data. Apache Flink [5] is a distributed data processing platform designed for real time data processing, providing a unified architecture for stream and batch data. In

the Flink process model, a program is mapped into a dataflow graph which defines the details of processing. JobManager coordinates the execution, delegating tasks to TaskManagers that are responsible for one or more operations. Regarding fault tolerance, Flink offers an exactly once guarantee using distributed consistent snapshots of operators and the current position of input stream. As resources of the authorities tend to be limited, winter service operations require careful planning balancing the costs and benefits. Deciding on an area to be cleared, route planning, distribution of routes across vehicles, and time requirements are responsibilities of the operation center. The algorithms developed for this problem can be classified into constructive, composite and metaheuristics categories, and are reviewed in [5]. From the computational point of view, the problem can be formally defined as a set of vehicles V with different capacities Vc, and a directed graph G = (N, A) with N representing set of nodes, A representing arcs with cost weights Aw, assigning paths to vehicles such that a postman tour on subset of G starting from depot, in which all edges are traversed with minimal total cost. This problem can be seen as a combination of the two well-studied problems of bin packing and the Chinese postman. In the standard bin packing problem, objects of different volumes are packed into bins in a way that minimizes the number of bins used or maximizes the total value of objects. The problem is NP-hard and algorithms that approximate to the optimal solution used to achieve a solution in polynomial time such as FirstFit and BestFit. FirstFit puts each item into first bin where it fits and BestFit puts into the most full bin where it fits and both have O(n log n) time complexity [6]. The Chinese postman problem is defined as finding the shortest closed path that visits every edge of a graph. The original problem was defined for undirected graphs with edges having no cost, and variants were developed for directed edges or other constraints. An important variant is the rural postman problem which can be applied to winter service operations in which a subset of edges are required to be traversed at a minimum cost. The rural postman problem has been shown to be NP-complete; therefore, heuristics were proposed for solving the problem efficiently [7]. In the route planning problem, the outputs of the Chinese postman routes become the objects for the bin packing problem; therefore, the capacities of the vehicles should be taken into consideration. III. SMART WINTER SERVICE PLATFORM An end-to-end platform is designed to ensure data obtained from road can be analyzed and visualized to improve winter service operations. Fig. 1 shows a diagram of the developed solution. The platform can be explained in three parts; data producer, data processor, and data user. In the data producer part, an electronic control unit is developed using 32 Bit ARM architecture microprocessor, which allows the use of GSM and GPS modules. The communication interfaces of the microprocessor are used with a sensor input. These include the amount of de-icing solution the vehicle is carrying, current rainfall, altitude,

slope, humidity, snowfall, air temperature, road temperature, and the speed of the vehicle. The electronic control unit contains GY-NEO6MV2 GPS Module, Load Cell, SeedStudio GSM/GPRS Module, Fuel Level Sensor, XS618B1PAL2 Inductive Speed Sensor, and ISS100 Infrared Temperature Sensors. The electronic control unit is also responsible for adjusting the salt spread according the vehicle speed, which ensures that the parameters for the amount of salt per square meter are satisfied. To the sensor data, the electronic control unit adds a timestamp and metadata related to the vehicle. The data is sent to the cloud via GPRS if a connection is available; otherwise, the data is stored in the electronic control unit and sent when a connection is available. In many situations, winter service vehicles work together on operations, and for that reason, the data processing architecture is designed to be scaled up according to the number of vehicles and the size of sensor data. Another feature of the data processing architecture is to be fault tolerant to ensure that in case of failure of a server, data will not be lost. Here, two key technologies are used; distributed message brokers and data processors. In distributed message brokers, data is replicated on different brokers to scale and increase fault tolerance. Message brokers are used to capture messages from winter service vehicles and distribute it to the consumers; i.e., data processors.

Figure 1. Diagram of Smart Winter Service Platform. The electronic control unit inside of the vehicle transmits GPS, sensor, and vehicle data to the cloud via GPRS, and it is visualized in web and mobile applications.

Data sent by winter service vehicles are acquired by Kafka message brokers on different topics which are used by the Flink data processors. This design makes it possible to analyze real-time streaming sensor data to compare environment of the vehicles and their status. Data is also stored in distributed databases for further batch analysis and reporting options in web and mobile applications, as well as for optimizing routing of the vehicles. A web application is developed to inform the operation center of the status of the winter service operation, which consists of a real time updated dynamic map, administrative screen, and reporting. In a dynamic map, the position of depots and responsibility areas, the real-time position of the vehicles, their routes, amounts of material, and work progress are visualized. This enables operation center personnel to manage all aspects of the winter service operation, from ascertaining the position of vehicles or any problems being experienced, as well as requests from citizens concerning road conditions. On the dynamic map, it is also possible to query operations in a given timeframe to compare results of a previous operation, or query a specific vehicle to analyze vehicle conditions and routes. This improves the transparency and liability of all operations, eliminating the potential of misconduct. A mobile application is developed for the Android system and designed to inform citizens about the winter service operations. After signing up, citizens can see the current position of vehicles, their routes and cleared roads on a dynamic map. Citizens can also take photos and create requests which pop up on the operation centers web application. With this feature, operation center is notified with previously unknown problems and can direct vehicle drivers to clear that area. This will help the public to plan their route, avoid poor road conditions, and reduce accidents. The novel features of the developed platform are as follows: • A multidisciplinary approach is adopted, in which embedded computing, IoT, geographic information systems, and big data processing are used. • All factors affecting winter operations performance are acquired and combined using IoT and cloud technologies. • Using the distributed cloud infrastructure, data can be processed fast and results in a system that is fault tolerant and operational 24/7. • Big data is visualized in a dynamic map to increase operation centers control over operation. Previous operations are also recorded and can be queried for any given time. • Members of the public can subscribe to the web application and receive notifications of the progress of the winter clearing operation and are able to choose to travel on roads that are safe. Citizens can also send messages to the operation center containing information or requests. The application of the developed platform can be extended to monitoring public transportation, construction

machines, cargo delivery, and monitoring condition of a vehicle for maintenance purposes. IV.

ROUTE PLANNING ALGORITHM

Planning winter service operations require different factors to be satisfied at the same time. These include reducing the costs of operation, optimally distributing the work across personnel and vehicles, clearing most of the roads, and doing it in a timely manner. The Smart Route Planning Algorithm allows planners to satisfy these conditions. In a general setting, the operation area is converted into a directed graph structure. In this structure, roads on the area are represented with edges, and intersection points are represented with vertices. Cost and benefit can be modeled as edge weights, and an objective function is required to maximize and minimize the distribution of routes. This process is shown in Fig. 2. In the second step, the graph is divided into subgraphs to assign possible routes to the closest depots. Subgraphs are obtained by detecting cycles in the main graph to provide a heuristic for route distribution. Here, the cycles are selected as a heuristic for routes because in each operation, winter service vehicles start from the depot and return to the same depot to refuel on completion of the operation. Therefore, optimal route for each vehicle are close to the cycles which contain depot node as one of their vertices. Route distribution is organized as a bin packing problem, in which there are objects with weights and containers with capacity. The goal can be defined as maximizing the total value of the containers. From the route distribution perspective, route weights can represent the cost of traversing a route, the amount of salt needed for that route or the time to travel the route and bin packing problem can be changed accordingly. This setting provides a flexible design in which the objectives of the operation center may change over time. In connection with SWSP, weather and road data can be used within the objective functions, enabling further

optimization possibilities. Segmentation of routes can also be undertaken considering the IoT sensor data, depending on average snowfall, rainfall, altitude, and slope of the road. As a greedy approach, edges not available in the initial graph partition can be added to the route of closest cycle without breaking the cycle. This makes it possible to satisfy travelling to all edges in practical applications, where clearing snow from all the roads is a requirement. Authorities responsible for winter service operations are usually equipped with vehicles that have different capacities. Also, in many cases personnel engage in manual clearing operations with a shovel for places that vehicles are not able to enter. The designed algorithm can be configured to distribute an equal work load among personnel, which can also reduce work distribution related problems, such as overworking. V. CONCLUSION Winter service operations are crucial for the safety of public transportation and citizens. For that reason, authorities try to plan efficient, fast, cost effective solutions with limited resources. In this work, a smart winter service platform is presented that increases the monitoring capabilities of the winter service operations, and a route planning algorithm is designed to distribute routes to the associated vehicles with an objective. The platform consists of an electronic control unit attached to the vehicle, sensors to obtain rainfall, altitude, slope, humidity, snowfall, air temperature, and road temperature, as well as the amount of solution the vehicle is carrying, distributed data processing and storage architecture, a web application for the operation center, and mobile application for public subscribers. Platform can also be used for other domains in which the real time monitoring of vehicles is required.

Figure 2. Diagram of Smart Route Planning Algorithm. Operation area is represented as graph, and cycles are decomposed to provide a heuristic for optimization

The Route Planning Algorithm is designed to enable the operation center to overcome different issues, to reduce the costs of operation, and to clearing most of the roads in an appropriate time. The algorithm converts an area into a directed graph and decomposes cycles in order to provide a heuristic for the possible routes. Depending on the edge weights and optimization functions, a solution is provided with the integer programming. This provides operation center flexibility in terms of satisfying different requirements which may change over time. Smart monitoring of the vehicles and their environment is an important aspect, and this can be extended to cover different purposes, such as in a vehicleto-everything setting in which winter service vehicles can communicate with other drivers to notify them of service vehicle positions and cleared roads, providing the drivers with safe transportation routes.

[2]

[3]

[4] [5]

[6]

[7]

REFERENCES [1]

L. D. Minsk, Snow and ice control manual for transportation facilities, New York: McGraw-Hill,1998.

[8]

S. S. Chakole, V. R. Kapur, and Y. A. Suryawanshi, "ARM hardware plaform for vehicular monitoring and tracking," Proc. 2013 International Conference on Communication Systems and Network Technologies (CSNT), IEEE Press, 2013, pp. 757-761, doi: 10.1109/CSNT.2013.162 P. Chen and S. Liu, "Intelligent vehicle monitoring system based on GPS, GSM and GIS," Proc. 2010 WASE International Conference on Information Engineering (ICIE), IEEE Press, 2010, pp. 38-40, doi: 10.1109/ICIE.2010.17 J. Kreps, N. Narkhede, and J. Rao, "Kafka: A distributed messaging system for log processing," Proc. of the NetDB, 2011 P. Carbone, et al., "Apache flink: Stream and batch processing in a single engine,” Bulletin of the IEEE Computer Society Technical Committee on Data Engineering, vol. 36, 2015 N. Perrier, A. Langevin, and J. F. Campbell, "A survey of models and algorithms for winter road maintenance. Part III: Vehicle routing and depot location for spreading," Computers & Operations Research, vol. 34, January 2007, pp. 211-257, doi:10.1016/j.cor.2005.05.007 D. S. Johnson, "Fast algorithms for bin packing," Journal of Computer and System Sciences, vol. 8, June 1974, pp. 272-314, doi: 10.1016/S0022-0000(74)80026-7 W. L. Pearn and T. C. Wu. "Algorithms for the rural postman problem," Computers & Operations Research, vol. 22, Oct. 1995, pp. 819-828, doi:10.1016/0305-0548(94)00070-O