Int. J. xxxxxxxxx xxxxxxxxxxxxxms, Vol. X, No. Y, xxxx 1 Title [ Do not insert author name or contact details now - only to be completed after acceptance ] Abstract: Keywords: 1 Introduction 2 Literature Review a. E-commerce In the past few years, e-commerce product delivery has grown dramatically. The majority of deliveries made in e-commerce (also known as "home shopping") involve packages, small items, and food [Manerba et al., 2018]. The world is experiencing a digital growth spurt, impacting almost every aspect of nature and human life. E-commerce and logistics integration will be a successful endeavour. Customer service, tracking, delivery, time effectiveness, and overall cost will all improve due to this incorporation [Keerthy et al., 2022]. Through the creation of new industries, fields of production, and services that were unthinkable even 30 years ago, the explosion of information technology services is profoundly reshaping the economy and the business world. E-commerce, in particular, has revolutionised consumer needs and behaviour, directly impacting how businesses conduct themselves [Archetti & Bertazzi, 2021]. The onset of the pandemic has sped up the process of consumers switching their purchasing behaviour from conventional brick-and-mortar retail stores to online e-commerce websites [Gupta et al., 2022]. The number of deliveries of goods to customers' homes increased as the e-commerce industry grew [Ranna et al., 2022]. Because they are closely related to customer satisfaction, quantity and quality are essential for the development of e-commerce in order to deliver products on time. Copyright © 202x Inderscience Enterprises Ltd. Author Therefore, in e-commerce, product delivery requires speed [EscuderoSantana et al., 2022]. Due to the customer's absence from the delivery location, there are an increasing number of first-time delivery failures as ecommerce continues to grow [Tilk et al., 2021]. Since ordered items must be delivered to specific customers, the popularity of e-commerce in today's world poses a significant challenge for last-mile delivery [Yuan et al., 2021] b. Last-Mile Delivery In order to reduce the cost of on-time delivery while maintaining quality, this research will concentrate on last-mile delivery in the e-commerce sector. The express or last-mile delivery market has grown due to the quickly developing e-commerce sector. Several activities and processes make up last-mile delivery from the last transit point to the delivery destination point, specifically at the B2C/end-user level[Ayu & Nahry, 2021]. The home delivery method is typically used for last-mile distribution in ecommerce businesses, where packages are delivered to customers' or end users' homes. In the e-commerce industry, last-mile delivery performance is crucial in the relationship between sellers and customers. This directly influences the customer's decision to reorder. Delivering goods to customers faster while lowering delivery costs is today's challenge for logistics companies offering last-mile delivery services. Contrarily, lastmile delivery is regarded as the most pricy, ineffective, and environmentally damaging link in the logistics chain. Last-mile deliveries account for 53% of total shipping costs [Ranna et al., 2022]. As a result, ecommerce is growing, and logistics firms must deal with an increase in parcel delivery, particularly in urban areas. Even though impressive ecommerce growth statistics have been provided, it has been noted that distribution costs, particularly for the last mile, could reach 40% of the price of a product [Mancini & Gansterer, 2021]. The customer must be present during the delivery process in attended home delivery, and delivery window requirements typically define the service. Deliveries may be made at a few collection locations (post offices, reception/delivery boxes, banks) or, more commonly, at the customer's home [Manerba et al., 2018]. According to research done, there can be two types of deliveries identified [Manerba et al., 2018] are; Fast delivery (FD) allows customers to place online orders up to a set deadline, after which they will receive their products in a specified window of time (usually a few hours). Such deadlines are set by retailers and are evenly spaced Title throughout the day (for example, six deadlines from 8 a.m. to 8 p.m., one every two hours). As a result, the vehicles of the retailers must leave the depot and return within a set number of time windows. This distribution strategy's speed is its vital strength. The retailer might be forced to use additional vehicles to serve customers within the allotted time. Moreover, another type is Lazy delivery (LD), The day before the delivery; customers must place their orders in advance. The customer can choose one of the time windows offered and will receive the delivery within that window the following day (Manerba et al., 2018). There are numerous last-mile delivery services available at the moment. Deliveries to homes are the most typical. Customers wait for their packages at home. Additionally, the delivery can be made to pick-up locations like stores or designated lockers [Yuan et al., 2021] The significant increase in requests for home deliveries has started to have a considerable impact on last-mile delivery. Because there are so many requests, businesses are forced to extend the delivery windows by several hours to guarantee that the delivery will be made during the customers' preferred time windows, typically at the beginning or end of the day. This could lead to a drop in the perceived quality of the service and, as a result, a decrease in customer satisfaction and loyalty. Additionally, efficiency losses are being experienced by logistics providers as a result of multiple visits to the exact locations caused by customers' absence. This harms ecological goals and transportation costs due to increased traffic in urban areas [Mancini & Gansterer, 2021]. Nevertheless, the ineffective distribution system for last-mile delivery is the issue with increasing the volume of parcels delivered to customers' homes. Due to this, more delivery vehicles are operating in cities, which can add to the negative externalities of traffic, noise, and pollution, all of which are harmful to human health. Failures during delivery to the customer's home or end user are another issue that frequently arises from last-mile delivery activities. Re-delivering the item to the customer's home will result in higher costs [Ranna et al., 2022]. The success of the online retail industry is heavily dependent on home delivery. However, last-mile delivery is currently regarded as one of the supply chain's most pricy, ineffective, and environmentally damaging parts [Manerba et al., 2018]. The rapid delivery of online orders creates new problems that the traditional routing problems of the previous 30 years did not have to deal with [Archetti & Bertazzi, 2021]. Author c. Vehicle Routing Problem with Time windows The logistics system problem of vehicle routing with time windows (VRPTW) has recently received much attention. The issue can be summarised as selecting routes for a small fleet of vehicles to serve a particular customer base within specific time intervals. Each vehicle has a specific volume. It begins at the depot and ends there as well. Each client should only receive one service. The VRPTW wants to keep the overall cost of transportation as low as possible. Researchers have focused their attention on the VRPTW, an essential variation of the VRP. The VRPTW aims to satisfy the time frame limitations and identify the best possible routes for a group of similar, restricted-capacity vehicles [Le et al., 2022]. The VRP with Time Windows (VRPTW), one of the wide varieties of VRP, is most often studied. Each request in VRPTW has a time window outlining the desired time frame for the service. Then, VRPTW seeks to minimise the violation of the requests' time restrictions when designing routes. Numerous efficient algorithms have been proposed to resolve VRPTW, and substantial investigations have been conducted[JacobsenGrocott et al., n.d.]. Recent years have seen much research into three different kinds of problem that considers dynamic customer wants: VRPTW with stochastic customer demands, VRPTW with SCDs and DCDs, and VRPTW with dynamic and stochastic customer demands. The subject of most studies is the VRPTW with DCDs in a single depot logistics network. (38) Numerous academics have researched the VRP extensively, notably the vehicle routing problem with time windows (VRPTW). When tackling difficulties, researchers frequently apply precise algorithms and metaheuristics. In recent years, some academics have also suggested specific techniques, like the branch-price and branch-and-cut algorithms, to solve VRPTW successfully. Exact algorithms provide a distinct benefit when solving issues of small size (fewer than 30 nodes), but when the scale increases, using these algorithms results in an unnecessary computational effort to solve the problem correctly and has a limited range of applications. Compared to the exact algorithm, the metaheuristic algorithm is better suited to solving complex issues since it conducts a more thorough search, which can increase the solution's efficacy and be used more broadly. The current approaches to overcoming these issues are genetic algorithms, ant colony algorithms, simulated annealing algorithms (SAA), and particle swarm algorithms [Xiang et al., n.d.]. Within the time between the earliest and latest timings, a vehicle serves each customer exactly once. Each vehicle journey begins and ends at the designated depot. Priority consumers' locations, demands, and time Title windows are known [Le et al., 2022]. An altered version of the Ant Colony Optimization utilising K-Means Clustering has been used in this research to solve the Vehicle Routing Problem with Time Window Constraint [Keerthy et al., 2022]. This research study [Manerba et al., 2018] examined how the length of time windows affects the distances the vehicles used in last-mile product delivery covers. The reduction of travel lengths (measured in kilometres) is a customary demand by retailers seeking to increase profits by cutting costs. To evaluate this, formulate these two policies regarding variations of the Vehicle Routing Problem (VRP). To be more exact, quick delivery uses a VRP with distance/time constraints (DCVRP), whereas lazy delivery uses a VRP with time windows (VRPTW). The proposed formulations for the two problems can be solved using a commercial Mixed Integer Linear Programming (MILP) solver since they are compact (i.e., contain a polynomial number of constraints). Then, test and analyse the issues using actual cases supplied by a shop that provided historical data for two days structured as in rapid delivery. The same real-case facts have been combined to create instances of the lazy delivery policy [Manerba et al., 2018]. In [Aydınalp & Özgen, 2022]mathematical model was constructed by adding time restrictions, capacity limitations, and maximum distance travelled to the models to create a mixed-integer programming (MIP) model based on the vehicle routing problem with time windows (VRPTW). As the model uses the GUROBI® solver to find the best solution for small issue sizes, metaheuristic techniques that imitate annealing and adaptive large neighbourhood search algorithms are proposed for solving significant problems. A real dataset was employed to evaluate the efficiency of the metaheuristic algorithms. The simulated annealing (SA) and adaptive extensive neighbourhood search (ALNS) were assessed and contrasted with GUROBI® and one another through a collection of actual problem examples. HFVRPTW (Heterogeneous Fleet Vehicle Routing Problem with Time Windows) refers to distribution system planning with many scenarios solved by vehicle routing issues considering various vehicle modes and time windows. VRP Spreadsheet Solver was used to address the VRP problem in this study [Ranna et al., 2022]. In addition to the customer Time Windows (VRPTW) variation, this research attempts to present a novel agent-based optimisation model to capture the uniqueness of vehicles in this routing problem. By centralising the solution search with a metaheuristic algorithm, methodology gets beyond the localised greedy solutions produced using decentralised and hybrid agent-based Author methodologies. Due to their rigidity in addressing combinatorial optimisation problems like VRP, heuristics evolved into metaheuristics, which is how centralised search has changed [Abu-Monshar & Al-Bazi, 2022]. In metropolitan locations, the Heterogeneous Fleet Vehicle Routing Problem with Time Window (HFVRP-TW) model is used to optimise the last-mile delivery route. The HFVRP-TW model is created and applied using information from express delivery during route optimisation. The information was obtained from an express delivery business, specifically the Jakarta, Indonesia, same-day delivery service, which was withheld for privacy reasons. The optimisation process was carried out through simulation and the creation of several operational scenarios. The necessary information comprises the location of the depot, the client's address, latitude and longitude coordinates, the kind of vehicle, the number of vehicles, the vehicle capacity, the service time, the earliest start time, and the latest end time (time window). Each customer must be served on or before the earliest time and before the latest time, which is known as the time window's two sides [Ayu & Nahry, 2021]. This research [da Silva Júnior et al., 2021]conducts a theoretical investigation, develops a novel method for a solution, and assesses the performance of optimisation algorithms (meta-heuristics) for the dynamic vehicle routing issue with time windows (DVRPTW). The two key contributions of this study are the framework for solving the DVRPTW and the proposed seven novel algorithms, as well as the successful outcomes in reducing the number of unsatisfied consumers. To solve the DVRPTW, had to offer seven algorithmic variations. Take into account both continuous and periodic re-optimisation. This research discusses the Electric VRP with Time Windows (EVRPTW), a variation of the well-known VRP with Time Windows (VRPTW). The EVRPTW manages a homogeneous fleet of EVs dispatched from a single depot to serve a group of clients with defined demands and time windows. The EVs depart from the depot to begin their routes, serving each customer exactly once before making their way back to the depot at the end of the day. They can be recharged at charging points along the way because they have a constrained driving range. The stations are few but need to be fully occupied. The goal is to travel as little as possible overall [Duman et al., 2021]. In order to accurately represent the nature of systems like the home visitation or home healthcare (HHC) systems, this study [Saksuriya & Likasiri, 2022] uses a heuristic for solving vehicle routing problems with time windows (VRPTW) with generic compatibility-matching between Title customer/patient and server/caretaker requirements. Any variant of VRPTW is more complex than standard VRP. Hence the solution requires a specific, tailored heuristic. The heuristic developed in this study is a successful combination of the inexperienced Local Search (LS), Ruin and Recreate method (R&R), and Particle Swarm Optimization (PSO). The suggested LS serves as the initial solution finder and the search engine for a practical/local optimal. The R&R component permits the solution to be over-optimised, and LS moves the solution back to the feasible side, while PSO aids in going from the current best solution to the next best solution. They solved 56 benchmark situations with 25, 50, and 100 clients to evaluate our heuristic and discovered that it could find 52, 21, and 18 ideal cases, respectively. Changed the benchmark instances to incorporate compatibility requirements to examine the effectiveness of our heuristic more thoroughly. The findings demonstrate that the heuristic can find the best answers in 5 out of 56 cases [Saksuriya & Likasiri, 2022]. d. Cluster-Based approach VRPTW can be addressed using a cluster-based algorithm to cluster constraints. There are few types of research done combined with vehicle routing problems with time windows. The objectives of this study are to establish the fleet size and reduce overall logistics expenses. Labor, gasoline and refrigeration costs are included in the logistical costs. The MIP model represents the 39-node product distribution network. The proposed MIP model is resolved using the clustering approach. According to the findings, 39 clients are served by 11 trucks, and the vehicle utilisation efficiency is greater than 70%. These ideal outcomes are discovered in an average of 0.36 seconds. Introduce the cluster-first-route-second strategy to address large-scale issues. This hierarchical strategy was applied on a big scale using a precise algorithm in a fair amount of time. A clustering algorithm is used in the first part of the procedure to create the clusters. Then, by modifying cluster members, each cluster's capacity is managed. In the second phase, the branch and bound algorithm is used to solve each subproblem for each cluster. Customers are grouped using the k-means clustering algorithm and the capacitated k-means clustering technique [Le et al., 2022]. In this study (Yuan et al., 2021), Instead of choosing just one delivery location while making an online purchase, a client can suggest a set of delivery destinations (home, pick-up spots, and car trunk) with the corresponding time limits. All these delivery services can be integrated and offered to customers. The courier must pick one of the addresses the customer has provided to deliver a product to a specific consumer. Author Customers gain from additional freedom following their preferences. It also lowers delivery costs and raises the percentage of successful first-time deliveries. The Generalized Vehicle Routing Problem with Time Windows can be used to mimic the underlying routing issue in the application discussed above (GVRPTW). The vertex set of a directed graph with clustered vertex sets is where the GVRPTW is defined. One cluster merely has the depot, where a uniform fleet of vehicles with fixed capacities is housed. The other clusters stand in for clients. The customer's potential delivery sites are represented by a cluster's vertices, each of which has a time frame during which the visit must occur if the vertex is visited. Each client has a demand attached to them. The goal is to identify a set of routes that minimises overall travel expenses, visits exactly one vertex per cluster, and abides by all capacity and time restrictions [Yuan et al., 2021]. In this research study [Dutta et al., 2022] proposed problem is to consider the closest cities or clients together in a cluster so that one truck can service them. This is done by using a cluster primary-route secondary strategy. As a result, it decreases the number of vehicles used and the carbon emissions those vehicles emit. The clusters are located using a modified distance-based k-means algorithm, and a collection of alternative pathways connected to each cluster is located using SPEA2. In order to choose the option that best fits the Decision Maker's preferences, apply the VIKOR approach. [Keerthy et al., 2022] propose transporting goods from the depot to the customer on time or earlier. This paper will explore how Ant Colony Optimization with K-Means Clustering (ACO-K-Means) has been used to save costs. The mathematical model described in this study will address the distribution, e-logistics, and retail network concerns. e. Literature overview This section summarises the literature on the VRPTW and includes a comprehensive descriptive analysis of papers published between 2017 and 2022. A list of research gaps that need to be filled and a literature summary round out the section. The VRPTW versions used in the prior literature are listed in Table 3.1, along with a synopsis of the problemsolving methods, objective functions, and kinds of algorithms employed. The selected articles were summarised after literature from the most recent six-year period, from 2017 to 2022, was analysed. Abbreviations for solution Methods: Title Meta Heuristic (MH), Heuristic (H), Mix Integer Programming (MIP), Mix Integer Linear Programming (MILP) Abbreviations for Characteristics: Vehicle Routing Problem with Time Windows (VRTW), Generalized Vehicle Routing Problem With Time Windows (GVRPTW), Travelling Salesman Problem With Time Windows For The Last Mile Delivery In Online Shopping (TSPTWLMDOS), Full Truckload Multi-Depot Vehicle Routing Problem With Time Windows (SFTMDVRPTW), Dynamic Vehicle Routing Problem with Time Windows (DVRPTW), Collaborative Multi depot Vehicle Routing Problem With Dynamic Customer Demands And Time Windows (CMVRPDCDTW), Open Vehicle Routing Problem (OVRP), Constrained Clustering for the Capacitated Vehicle Routing Problem (CC-CVRP), Electric Location Routing Problem With Time Windows (E-LR Table 2.1. Literature Review Summary Table Author Reference [Wu & Wu, 2022] [Aydınalp & Özgen, 2022] [Gupta et al., 2022] [Ranna et al., 2022] Solution Method characteristics Fleet of vehicles Depot Product MH VRPTW Homogenous S M MH VRPTW Homogenous S M VRPTW Heterogenous S M VRPTW Heterogenous S M VRPTW Heterogenous M M VRPTW Homogenous S M GVRPTW Homogenous S M VRPTW Homogenous S M VRPTW Heterogenous S M VRPTW Heterogenous S M TSPTWLMDOS Heterogenous M M SFTMDVRPTW Homogenous M M DVRPTW Homogenous S M CMVRPDCDTW Heterogenous M M OVRP Homogenous M M CC-CVRP Homogenous S M VRPTW Homogenous S M E-LRPTW Homogenous S M VRPTW Homogenous S M Deep-RL model VRP Spreadsheet Solver [Dondo & Cerdá, H 2007] [Escudero-Santana MH et al., 2022] [Yuan et al., 2021] H / cluster MIP / [Le et al., 2022] cluster [Abu-Monshar & agent-based Al-Bazi, 2022] MH VRP [Ayu & Nahry, Spreadsheet 2021] Solver [Jiang et al., 2020] H [el Bouyahyiouy MIP/ & Bellabdaoui, Generic n.d.] algorithm [da Silva Júnior et H al., 2021] [Wang et al., IMOPSO2022] DIS Cluster[Dutta et al., based 2022] algorithm Cluster[Alesiani et al., based 2022] algorithm MIP/ K[Keerthy et al., Means 2022] Clustering MILP/ K[Sánchez et al., Means 2022] Clustering [Villalba & la H / k-means Rotta, 2022] clustering Title [Manerba et al., 2018] MILP VRPTW Homogenous S M Proposed Study H/ Clustering Algorithm VRPTW (based on Receiver) Heterogenous S M Abbreviations for Deport and Product: Single(S) , Multiple(M) After analysing all literature, the Vehicle routing problem with time windows-based papers used mostly meta-heuristic or heuristic-based algorithms to solve VRPTW. According to past literature cluster-based approach to VRPTW is limited and only used for unique scenarios, and most of every paper looked at VRPTW from the senders’ point of view and applied solution but not the point of view of the Receiver. Table 2.2. Literature review summary on objectives Reference [Wu & Wu, 2022] [Aydınalp & Özgen, 2022] [Gupta et al., 2022] [Ranna et al., 2022] [Dondo & Cerdá, 2007] [Escudero-Santana et al., 2022] [Yuan et al., 2021] [Le et al., 2022] [Abu-Monshar & AlBazi, 2022] [Ayu & Nahry, 2021] [Jiang et al., 2020] [el Bouyahyiouy & Bellabdaoui, n.d.] [da Silva Júnior et al., CM X Objective Functions CS OLM RNV RTD X X X X X X MCE Industry Agri – Ecommerce Pharmaceutical Industry E-commerce E-commerce Logistic X E-commerce X X E-commerce Logistic X X Logistic X E-commerce Parcel Delivery X X X X Truck Delivery E-commerce Author 2021] [Wang et al., 2022] [Dutta et al., 2022] X X X X [Alesiani et al., 2022] [Keerthy et al., 2022] X X [Sánchez et al., 2022] [Villalba & la Rotta, 2022] [Manerba et al., 2018] Proposed Study X X X X X Logistic E-commerce Postal ecommerce Transportation Electric Vehicle Transport X Logistic X X E-commerce E-commerce As of Table 2.2, after analysing objective functions and Industry, most papers done cost minimisation and minimisation of carbon emission but not for the VRPTW from the viewpoint of the Receiver but from the sender. Moreover, Last-mile delivery in the e-commerce sector has been affected chiefly in the last few years due to covid-19 and the economic crisis. However, it had both positive and negative impacts on last-mile delivery. Considering all those factors, we have selected VRPTW from the point of view of receiver address using heuristic and cluster-based algorithms to Heterogenous vehicle fleets, single deport and Multiple products aiming to reduce travel distance and cost for last-mile delivery in the e-commerce industry. Abbreviations for Objective Functions: Cost minimisation (CM), Customer Satisfaction (CS), Reduce Number of Vehicles (RNV), Reduce Travel Distance (RTD), and Minimise Carbon Emission (MCE) Title 3 Methodology 4 Basic model 5