Bees algorithm for Forest transportation planning optimization in Malaysia

Abstract

Algorithm is widely used in various areas due to its ability to solve classes of problems. Due to multiple objectives to be met and varied algorithm application in this digital era, addressing the problem-solving optimization in a more efficient and effective way has become more reasonable. Forest transportation planning is one of the most expensive activities in timber harvesting and can be optimized through algorithm application. Forest transportation planning is a vital component of timber harvesting activities. Inappropriate planning may raise the overall costs of harvesting activities. This paper aims to give an overview of several algorithm application in optimizing the forest transportation planning problem and give an insightful information regarding the relationships between algorithm and the integration of transportation system characteristics and variables. Examples of algorithm that are finding their way to the forest transportation planning problem include Genetic Algorithm (GA), Particle Swarm Optimization (PSO) algorithm, Ant Colony Optimization (ACO) algorithm, Simulated Annealing (SA) algorithm and Tabu Search (TS) algorithm. Although no literature was found regarding forest transportation planning problem optimization with regards to Bees Algorithm (BA), rules set for several transportation problem evidenced from literature search seems to be applicable to forestry. Generally, in this paper, the BA has been given focus for forest transportation planning problem optimization as a potential algorithm to overcome the challenges of environmental degradation and efficiency of timber extraction used, as well as its accuracy and less processing time for problem-solving.

Keywords:

• Operation research
• optimization
• forest transportation planning
• timber extraction
• linear programming

1. General background

Forest transportation is part of wood supply chain (Frisk et al. 2010). It is particularly important in timber harvesting planning as discussed by several researchers such as Anderson and Nelson (2004), Mohd Hasmadi and Kamaruzaman (2009), Chung and Contreras (2011), Norizah et al. (2014) and Dai et al. (2017). Trucks are typically used in forest transportation that serve as a mean to transfer loadings such as logs from extraction area either directly to customers or indirectly to final destinations such as landings or mills (Epstein et al. 2007). According to Carlsson and Rönnqvist (1998), Carlsson and Rönnqvist (2007), Forsberg et al. (2005), Mohd Hasmadi and Kamaruzaman (2009) and Norizah et al. (2014), transportation planning in forestry can be described into three major planning, namely; (i) strategic, (ii) tactical, and (iii) operational. Decision made to implement strategic planning is influenced several factors including efficiency of harvesting activities, well-maintained road structures, and capability of transportation modes. The requirements mentioned deem strategic planning to be the most expensive in comparison to the other (two) transportation schemes (Chung and Sessions 2003; Contreras and Chung 2007; Barros et al. 2010; Frisk et al. 2010; Norizah and Chung 2014). On the other hand, tactical planning only requires short-term arrangement provided that details related to the operation (e.g. extensive spatial information of the scheduled processing area) are prepared. Meanwhile, operational planning is an ad-hoc decision made during harvesting operation to ensure in-time delivery of timber products to customers (Wilhelm 1999) and well planning in timber harvesting is important in timber harvesting in order to sustain the timber product (Eshun et al. 2010).

Nevertheless, forest transportation planning is not only limited to timber production. For example, Kar and Jacobson (2012), Zamora-Cristales and Sessions (2013), Zamora-Cristales et al. (2015), Zamora-Cristales et al. (2017), Chen et al. (2017), Malladi et al. (2018), Akhtari et al. (2018) and Álvarez-Miranda et al. (2019) put an effort to optimize forest transportation to relocate biomass to be utilized as a source of energy at a demand area. Evans (2007) and Naik et al. (2010) clarified that energy from biomass is important to create biofuel energy and biofuel product. The similarity between forest transportation and biomass transportation lies in the planning stage that depends on the need to reduce wastage typically incurred from costs or time of deliverance.

In addition, forest transportation also corresponds to a large proportion of the total operational cost, which is a significant measure of efficiency in timber harvesting operation (Epstein et al. 2007). Mohd Hasmadi (2009) stated that although excessive length of road construction contributes to the high cost of forest area opening, well-organized forest transportation planning could reduce the overall cost of harvesting activities. Designing forest transportation road is a complex procedure that requires details and meticulous planning. Studies on designing and planning for forest transportation have been conducted by several researchers; among other, Broad (1985) applied linear programming to optimize forest transportation and management system. The algorithm links the networks as advantageous which in turn improves the phase of forest transportation. Meanwhile, Equi et al. (1997) used Simulated Annealing (SA) algorithm to reduce transportation cost and scheduling problem. They combined both problems and used the algorithm to appraise alternatives that minimize spending and coordinating schedule suitable for forest transportation. Chung and Session (2003) designed a cost-saving road construction while providing alternative routes for transportation activities by using optimization method to solve transportation problems. Similarly, Contreras and Chung (2007) solved least cost forest transportation planning problem using the optimization method in northern Idaho with environmental impacts as a problem constraint in their study. Another study by Norizah and Chung (2014) in the tropical forest of Peninsular Malaysia had determined the minimum cost timber harvest system between two types of skidding activities called conventional (Crawler Tractor- CT) and reduced impact logging (Log Fisher- LF) skidder by using the optimization method through considering several engineering specification criteria as problem constraints.

The studies discussed above have shown several applications of the optimization methods in designing least cost forest transportation planning. Local optimal solution would be feasible if several factors of objective function and problem constraints were considered. Eriksson and Björheden (1989), Flisberg et al. (2009), Frisk et al. (2010), Malladi and Sowlati (2017), and Yan and Ryoo (2018) suggested that optimization should be applied either through linear programming or multi-linear programming. According to Soyster (1973) and Karmarkar (1984), linear programming is a technique used to derive optimization of a linear objective function, subject to linear equality and linear inequality constraints that are based on repeated applications of such projective transformation. It utilizes optimization over the inscribed sphere to create a sequence of points which converge to the optimal solution in order to solve the problem (Mohammadi et al. 2017; Ebeniro and Osho 2018). The feasible region generated is a convex polytope, which is a set defined as the intersection of finitely many half spaces, each of which is defined by a linear inequality. Its objective function is a real-valued affine (linear) function defined on this linear programming. A linear programming finds a point in the polyhedron in which the function has the smallest or largest value if such point exists. A number of researchers have employed linear programming for optimization purpose in the forestry field, and this is usually for cost optimization and planning such as in the studies by Augustynczik et al. (2016), Acuna (2017), Martin et al. (2017), and Mohammadi et al. (2017). A study by Augustynczik et al. (2016), applied linear programming to aggregate harvesting activities in forest plantations. They formulated harvesting scheduling problems with aggregation of harvesting activities requirements and included biodiversity concerns into harvesting planning. Linear programming will be able to maximize the cluster of harvesting areas with consideration to biodiversity concerns. Similarly, Acuna (2017) used linear programming to solve complex transportation planning problems involving annual, monthly and daily decisions for timber and biomass transport. The linear programming had optimized the flow of timber and biomass from forests to mills and energy plants, and the efficient routing of trucks that transport those products for the transport planners from the Australian forest industry. Martin et al. (2017) applied linear programming for timber harvesting in Woodland Caribou, Canada to maximize timber harvesting production while minimizing a proxy measure for wildlife habitat. They were able to widen the area of wildlife habitat alongside with timber harvesting. They found that their model could be a useful tool to guide managers when facing conservation and restoration of old-growth forest stands alongside industrial timber production. Besides, Mohammadi et al. (2017) had used linear programming to determine the ecological capabilities classification of different land species for plantation for optimal forest plantation. Their model could determine two species that can be economically profitable for plantation. Their result was beneficial for the forest managers to consider economic and ecological criteria in forest plantation.

In contrast, multiple programming objective functions and constraints are treated as variables and problems, while linear programming generalizes the problem in a straightforward manner (Drenick 1992; De Campos and Cozman 2007; Lukatskii and Fedorova 2017; Charkhgard et al. 2018). A study by Sokouti and Nikkami (2017) optimized land use conversion from forest land in reducing soil erosion by using multiple programming. On the other hand, the study by Mendoza and Prabhu (2000) applied multiple programming as a decision tool to assess criteria and indicators in evaluating sustainable forest management. Their study concluded that the calculation in each evaluation step can be easily understood by the participants.

1.1. Significance of Forest transportation studies

It is crucial to note that different countries implement different practices of forest operation system and this leads to differences in transportation planning. In some countries, heli-logging is used due to suitable features such as less dense forest landscape like in several temperate countries of Columbia (Roberts et al. 2004), East America (Wang et al. 2005), Alaska (Christian and Brackley 2007) and Turkey (Akay et al. 2008). Ruefenacht et al. (2008) suggested that sparse forest canopy makes heli-logging more suitable to be utilized in temperate forest compared to tropical forest that has denser vegetation structure. Lower cost is expected from heli-logging in transportation planning due to the smaller road network to be created within timber harvesting area.

In a different situation, skyline harvesting system might also lessen the forest transportation planning problem as only corridor is established without earthwork to assist timber extraction activity. Skyline harvesting system is more practical in a hill topography. This system uses suspended steel cable to transport logs from harvesting site to landing. The logs may be fully or partially suspended for the entire or a portion of the skyline distance. The skyline’s cable loop runs around a drive pulley, generally at the central delivery end, while the return pulley is located at the collection end. This system may operate over a large area and is being widely practiced in several areas such as in Northern California, United State of America (USA) (Hwang et al. 2018), Lushoto Tanzania (Dos Santos 2015), Italy (Proto and Zimbalatti 2016), and France (Spinelli et al. 2017). In tropical forests, this skyline harvesting system found to be useful and more practical in Indonesia (Endom and Astana 2017). A study conducted in 2011 and 2013 by Endom and Astana (2017) found skyline logging as effective at steep slopes ranging from 40% to 60% in terms of timber volumes. In Malaysia, however, there are limitations to use heli-logging and skyline harvesting systems in forest harvesting due to higher operational costs and denser tropical stand structures (Pinard et al. 1995; Bigsby and Ling 2013). However, there are some reports stating heli-logging is being utilized in Sarawak, and to a lesser extent, in Sabah (Thang 2004). Bigsby and Ling (2013) found significance of heli-logging from long-term productivity by higher number of turns per hour made to relocate logs in Sarawak. They stated that heli-logging has the potential to become an important part of the mix of low-impact harvesting systems that are necessary for sustainable forestry in the tropics. However, their significantly higher operating cost means that the plan in maximizing productivity is critical for this system.

1.1.1. Forest transportation in Malaysian Forest

Advances in forest transportation planning has seen technology modified ground CT to be less impactful to the earth surfaces (Chung et al. 2008; Jourgholami 2012; Tavankar et al. 2015; Breschan et al. 2017; Proto et al. 2018). Their studies tested modified ground CT for timber extraction to reduce impact on residual trees, minimize impact to the road surfaces, and capability of such machineries to operate at steep slope areas. Consequently, ground CT is deemed as more practical to be applied on undulating topography with hill steep slopes of Malaysia. Since 1942, all activities in timber harvesting in Malaysia have been mechanized (i.e. CT and truck tractor) which timber harvesting operation found to be faster and rapid (Kamaruzaman and Shah Nik Mustafa 1994; FDPM 1999, FDPM 2001; 2003). This can be traced back to the Japanese occupation era (1942–1945) where many forest areas were converted to grow food resources as evidenced by wide canopy opening (FDPM 2011; Norizah et al. 2011).

CT is a machinery used to skid log from felling site to temporary landing. Prior to that, CT is used to excavate earth surface for road construction (i.e. feeder road) and clear trees to create skid trail to reach the felling site. In Peninsular Malaysia, truck tractor is also known as “San Tai Wong” (Norizah et al. 2016), which is then proceeded by transportation logs from landing at the harvesting site to primary landing located outside the timber harvest area. This transportation system is a common extraction technique used in Peninsular Malaysia. In 1988, Standard Road Specification was launched by FDPM to ensure proper forest road construction (i.e. Class I, Class II, Class III and Class IV) to accommodate mechanized timber harvesting operation (FDPM. 1988). As there are frequent passes of CTs in the same skid trail, greater impact to the environment happens during timber harvesting operation (Wyatt-Smith 1954) and after timber harvesting operation (Razali et al. 2014). The studies by Kamaruzaman (1988), Kamaruzaman (1991), Kamaruzaman and Nik Mohamad (1992), and Mohd Hasmadi and Norizah (2010) supported the statement expressed by Wyatt-Smith et al. (1954) due to several characteristics. Kamaruzaman (1988), in his study, found that soil compaction is significantly higher on CT tracks. He suggested compacted soil creates resistance for water penetrate through. A different study by Kamaruzaman (1991) reported that continuous timber harvesting operation, regardless of the weather conditions, is potentially more damaging to site productivity. In the same study, greater changes in soil density and porosity were also found from operations that utilized wheeled skidding machine. Meanwhile, Kamaruzaman and Nik Mohamad (1992) suggested that timber harvesting operations conducted in hill forest had caused severe disturbance to the soil where bulk density were found to be higher on a skidding track. A comparison done on the effects of mechanized harvesting between two machineries (CT and LF) showed that the latter causes higher significant impact to soil disturbance (Mohd Hasmadi and Norizah 2010).

Norizah et al. (2011) suggested that the use of machineries as key drivers for soil disturbance in timber harvesting operation. In return, FDPM has shown the efforts to fight the negative impacts of timber harvesting operation by amending several guidelines for forest roads. The new guideline released in 2010 was an amendment made to improve the 1999 guidelines that had focused on Class III and Class IV forest roads. In the 2010 guideline, the long cable system known as LF was introduced as a new mode of timber extraction, but no specific guideline focusses on LF. In 2013, amendments were made to the 2010 road guideline, which particularly aimed to improve skid trail specifications. Skid trail is a trail connecting feeder road from stump area to log landing. Forest transportation with LF then ruled under specific guideline published in 2016. To this date, not more than 5 licensees (the contractors who were given permission and license to do timber harvesting operation by the State Forest Director) are conducting timber extraction through LF in Peninsular Malaysia. LF is a 300-metre-long haulage cable system attached at the long boom operated with a hydraulic system. The extraction method of this machinery was adapted from the fishing rod where logs are tied with a choker cable and let to be partially suspended with the end part of the logs touches the ground during the hauling process. Log fishing ends when the logs are placed horizontally on the ground surface at an open area such as at the temporary log landing or LF platform (FDPM 2016). Specifications for the LF facilities were released by FDPM in 2016 to ensure proper forest transportation planning system and construction. This extraction technique requires no skid trail creation and it is known as one of the Reduced Impact Logging (RIL) practices that is believed to minimize impacts on soil. However, studies that examined the impacts of LF from timber harvesting operation are limited (Gan et al. 2006; Jusoff and Taha 2008; Abdul Rahim et al. 2009; Norizah et al. 2011; 2012; Norizah and Chung 2014; Lim and Yusof 2017; Noraida et al. 2017), while the effectiveness of LF was only highlighted in the guidelines released in 2016. Figure 1 illustrates the timber harvesting activities using both CT and LF in timber extraction activity.

Figure 1. An illustration of timber extraction using CT and LF from stump site to landing.

Timber extraction techniques may incur higher operational costs as opposed to other techniques (As et al. 2009; Medjibe and Putz 2012). Incurs on operational cost may be due to allocation made to create suitable transportation plan and system that are justified with the capability of machineries to efficiently conduct log extraction activities, although this does not exempt its significant impacts on the forest environment. Employment of appropriate machineries and timber extraction techniques is important to lessen the impacts on the environment. Studies by Mizoue et al. (2016) suggested skid trail planning, pre-harvest liana cutting, and directional felling can facilitate timber extraction and protect timbers from damage and lower the price. However, Khai et al. (2016) stressed that road construction may trigger problems related to soil disturbance. These issues need to be addressed during strategic and tactical planning to ensure effective forest transportation planning while minimizing impacts on the environment. Since 2000, a number of reports have yielded some reliable approaches to properly plan forest transportation system in Malaysia by considering several constraints. For example, Mohd Hasmadi and Kamaruzaman (2009) used the GIS-Best Path Modeling to design access road plan for timber harvesting operation to have an optimum road density as required by Forestry Department of Peninsular Malaysia. The study conducted by Norizah and Chung (2014) used the graph theory to design timber harvest area planning for two skidding techniques (between CT and LF) with emphasis on cost minimization on the overall forest transportation planning system. Their study considered all the specifications listed in the forest road guidelines by FDPM (FDPM 1999, 2013, 2016) while loosening some restrictions posed from one or more regulations. Planning for the least cost forest transportation is still a crucial subject to address in Malaysian forestry. This is because planning for a forest road network requires a thorough consideration of several regulations addressed in the forest road guidelines (FDPM 1999, 2013, 2016).

These article reviews several optimization methods to solve forest transportation problems previously studied worldwide and their potential applications to be applied to Malaysian forest. In addition, our focus, bees’ algorithms, the first to be applied in this kind of problem due to its better quality of their solution and the good calculation time performance, was given extra attention in Section 4.0. The strength and limitation of BA in transportation problems from previous application was also discussed with the arguments of solutions concepts the past application tried to elucidate were similar with Malaysian problems.

2. Materials and methods

The literatures and information for this review paper were obtained from (1) published papers, (2) book, (3) reports, and (3) bulletins by searching electronic databases including Google Scholar, Scopus and Web of Science, and non-electronic database including Guidelines. This review was conducted through semi-systematic review. Snyder (2019) describes this type of literature review approach is designed for topics that have been conceptualized differently and discussed in any disciplines. The keywords used in the search included; forest harvesting, forest harvesting machines, timber extraction, Malaysian forest, impacts on forest roads, optimization algorithm, operation research and bee’s algorithm. We used 129 studies found in the literature review from 1954 to 2020 for this manuscript. Of that, 94 studies were related to forest transportation in Malaysia which are integrated with timber harvesting planning, and 35 studies were related to operations research models and method of optimization algorithm to support transportation decision planning.

3. Timber transportation with optimisation algorithm

In transportation activity, decisions on road alignments are made on multi-objectives considering different constraints. Thus, when more than one objective and constraints are confronted, many considerations need to be addressed in the same time. This can be described as multidimensional and mathematical problem (Behera et al. 2015). In optimizing the objective function while minimizing the constraints, automated processing offered by algorithm optimization could fasten the decision making. The capability of algorithms in solving problem has become the focal point in the forestry field especially in forest transportation. The following literature is focusing on algorithm in forest transportation planning.

In forest transportation, the Genetic Algorithm (GA) has been used by Mullen and Butler (2000) for forest operational planning and harvest scheduling issues. They found GA with a combination with heuristic scheduling algorithm is more efficient techniques to optimize the timber harvest scheduling problem as compared to linear programming. The study by Chung and Sessions (2001) highlighted the GA as a tool that could suffice designing of forest road network by solving road routing problems. In their study, the combination of GA with SA was done to optimize road transportation problem. However, they stated that the GA technique implemented in their application might have limitations in generating new solutions because the processes of random crossover and mutation could be restrictive. On the other hand, Aruga et al. (2005) applied the GA in their study for development of an automated forest road design programme as a way to minimize costs incurred by construction and maintenance, along with evaluation of sedimentation from soil erosion on forest roads. Their aim was to provide optimum cost of forest transportation infrastructure for CTs and forwarder used in timber harvesting operation. They used the GA and Tabu Search (TS) to compare the best optimum cost and found that the GA provided the most satisfactory solution and method to reduce cost and time of forest transportation compared to TS.

Similarly, study by Aruga (2005) optimized the horizontal and vertical alignments of forest road construction by using the TS algorithm. In their study, the TS algorithm optimized forest road alignments based on the total costs of forest road construction. Using the high-resolution digital elevation model (DEM) derived from light detection and ranging (LiDAR) data, they generated ground profile and cross sections for the TS algorithm to define the optimal horizontal and vertical forest roads. The TS algorithm was able to define the suitable alignment for the forest roads and their study helped forest engineers to design forest roads by evaluating a large number of alternatives.

Meanwhile, Fotakis et al. (2012) availed GA with additional spatial operator and found it to refine efficiency for multi-objective spatial forest planning compared to GA without any additional spatial operator. Other than that, the Particle Swarm Optimization (PSO) algorithm was also used by Izquierdo et al. (2008) to reduce the cost of operational plan for forest biomass transportation from the landing point to the energy production plants. Their result showcased the PSO algorithm as capable of reducing the operational cost from solutions to the strategic planning for optimal biomass supply chain starting from harvesting to transference phase. Furthermore, the PSO algorithm has also been used by Shabani et al. (2013) in providing the optimum solution for decisions related to network design, technology choice, plant size and location for forest biomass chain. Their study found it to be able to optimize and improve the design, management and performance of a forest bioenergy supply chain.

Besides, the Ant Colony Optimization (ACO) algorithm was used by Chung and Session (2003) to find the shortest-path for forest transportation plan with CT harvesting system, and it was found to be able to solve their problem by counting the neighbor node for road network planning with optimum cost of production. The results indicated that solving transportation problems with multiple goals enables users to analyze tradeoffs between goals and generate alternative road networks that can reduce both transportation costs and negative environmental impacts. Similarly, Contreras et al. (2008) applied the ACO algorithm by using fixed and variable costs without ignoring the side constraint of total sediment yields from the road network construction for a ground-based harvesting system consisting of CT and forwarder machine in which they were able to solve large and complex forest transportation problems.

Other researchers who have used the ACO algorithm in forest transportation planning problem for a ground-based harvesting system were Lin et al. (2017). Their study showed that the ACO algorithm can produce near-optimal solutions for all constraints in forest transportation planning problem compared to the mixed-integer programming (MIP) which has been commonly used to curb with the same issue. They added that the ACO algorithm needs improvement to ensure quality and time efficiency for larger and more complex problems. They also suggested that the ACO algorithm should focus on time efficient technique to fine tune parameter values without the need to conduct an exhaustive parameter search.

In a different study by Haridass et al. (2014), the SA algorithm was used in forest transportation planning with forwarder to solve scheduling problem under the capacity and time window constraints. In particular, the SA algorithm was employed to design the best possible route for timber transference to mills. Interestingly, the SA algorithm yielded feasible solution and outperformed manually generated results. Chung and Session (2000) also used the SA algorithm to solve forest transportation planning during harvesting by optimizing the fixed and variable costs. The SA algorithm was developed with the main aims to reduce road construction cost and provide alternatives to transportation network when CT was used in timber harvesting operation. A study by Aruga et al. (2005) planned the heuristic technique in the development of an automated forest road design for a timber harvesting operation with CT to minimize construction and maintenance costs, and evaluate sedimentation from soil erosion on forest roads by using GA. They used costs of construction materials as constraints. Additionally, they compared the GA with TS algorithm, and found that GA provides better solutions than TS algorithm.

Zamora-Cristales et al. (2015) used the ACO algorithm to plan the least cost for forest biomass processing and transportation network. Their study appraised lower cost than the usual to yield the optimum forest road network for biomass transportation which was collected from the log residues in a landing area. The ACO algorithm provides decision support at the operational level to develop alternative routes to optimize biomass processing and transportation from the harvesting area.

Meanwhile, Gracia et al. (2014) applied the GA method for vehicle routing problem in biomass transportation from forest. They used this algorithm to optimize the vehicle routing problem by minimizing the cost of transportation network. They stated that this algorithm had improved and optimized the vehicle routing problem as well as generated collective sequences of residual biomass which minimized the total traveled distance in and out of the field.

Other than that, Richards and Gunn (2003) used the TS algorithm in solving the stand harvesting and road access optimization problem in forest transportation planning for a combination of skyline and ground-based harvesting system; CT. In particular, they used the TS algorithm to optimize stand harvest and road construction schedules, in which different search criteria of the TS algorithm were utilized for random diversification moves, and as a result, fixed functional prevailed as a better result that produced feasible solutions.

One of the promising algorithms not yet applied to timber transportation is the Bees Algorithm (BA), an optimization technique introduced in 2005 by Pham and colleague (Pham et al. 2005). To date there have been numerous applications of BA explored to solve several different optimization problems.

4. Bees algorithm in timber transportation

BA is an optimization algorithm inspired by the natural behavior of honey bees to find an optimal solution. BA avoids getting trapped in a local minimum solution because of its capability of performing in both local and global search. The description of BA is given in Equation (1)(1) $$v_{ij}=x_{ij}+{\mathrm{\varnothing }}_{ij}\left(x_{ij}-x_{kj}\right)$$(1) . $\in \left\{1,2,\dots .\mathrm{}SN\right\}{\mathrm{\varnothing }}_{ij}.$ (1) $$v_{ij}=x_{ij}+{\mathrm{\varnothing }}_{ij}\left(x_{ij}-x_{kj}\right)$$(1)

Where k and j are determined randomly $\in \left\{1,2,\dots .\mathrm{}SN\right\}.$ The values of k and j must be different from i, and ${\mathrm{\varnothing }}_{ij}$ is a random number between (–1, 1) and v represents the fitness.

BA is widely applied in various fields such as mathematical, engineering, and social sciences (Chong et al. 2006; Kavousi et al. 2012; Mansouri et al. 2015). Nonetheless, the application of the BA in forest transportation problem is rather limited in the literature. Yet, there are several studies on the utilization of BA in urban transportation planning.

Iqbal et al. (2015) used BA and TS to solve vehicle routing problem with multiple-objective functions. Solution model with a BA was able to solve the problem with balanced exploration and exploitation to avoid local optimal and reach the global optimal. Their model considered minimizing the cost of total traveling distance with penalties, the number of vehicles routes and number of window violations. In this study, the BA approach was quite effective, as it provided a high-quality solution within reasonable computational time. As compared to TS algorithm, BA provides a high-quality solution, confirming the traveling distance is shorter by 1005.21 m as compared to 1388.00 m with TS algorithm. In terms of processing speed for computational time, solution with BA only taken 5.3 to 13.55 seconds as compared to TS algorithm which took 52 to 82 seconds. But, the limitation of their study was this model was unable to consider both fixed and variable costs with the time window. Similarly, Ng et al. (2017) applied BA for a capacitated vehicle routing problem and re-routing strategies under time-dependent traffic congestion. They developed a model to examine through capacitated vehicle routing instances and the results indicate the effectiveness and the potential of using real-time information for data-driven vehicle scheduling. After some modifications made to their bees’ colonies, the performance of solution increased by 13.52% with 12.57% deviation from best known solution. The limitation in their study was unable to consider real-time multiple-objective functions such as weather condition and road constructions.

Zhang et al. (2017) had implemented BA with TS to vehicle routing problem with realistic constraints in the logistics industry. Their model was able to consider time window constraints and closely reflected real-world situations. The BA provided an efficient solution in solving the vehicle routing problem with realistic constraint but less efficient for the constraint that was included for the initial loading in the starting points. George and Binu (2018) tested BA and Cuckoo Search Algorithm (CSA) to solve a complex problem between logistics service provider with the customers. In order to satisfy the customer needs, and to overcome the restriction on the availability of materials and transportation resources, they plan the solution with multi-objective functions. They used a different number of population size of bees and birds between 100 and 500 for both algorithms. BA was tested as the best solutions for their problems than CSA with 99.55% difference in terms of multi-modal test function. Since their study used about 500 bees and cuckoo respectively, for BA and CSA, the solution’s time increased and it is expected that high speed computer processor is required. Chen and Zhou (2018) in their study revealed when the number of iterations for BA was set too high, the solution speed will be slow and it was easy to fall into the local optimum, thus unable to specify the constraints of the problem.

Karaoglan et al. (2020) used BA with the distance-constraints for vehicle routing problem. They used BA to minimize the cost of transportation operations from 19 towns by the routes with a maximum length of 550 km. The BA was able to solve the distance restricted vehicle routing problems with the number of the routes reduced from 6 to 4. They also compare their results of BA solution model with the results from the common practice for vehicle routing problem in Turkey and found the total distance reduced from 1452 km to 1031 km. However, their model was unable to consider real-life characteristics such as multiple visiting locations.

Meanwhile Santosh and Suresh (2019) used BA with the traveling time of a student in a bus-constraints for school buses traveling problem. In their study, the long traveling time of a student in a bus is considered as constraints in order to increase the operational efficiency and reduce the overall cost and routes to travel. Five routes were taken in the experiment with time-traveling were taken for normal operation and implemented the BA. The result for five routes in normal operation and after BA was applied decreased from 155.13 hours to 146 hours of traveling time. The operational efficiency of school bus traveling increased to 18.67% from standard routing routes after solution models with BA. Since they unable to consider multi-objective, including traffic time, maintenance, breakage, and emergency conditions in their optimization, hence, this has become a limitation in their study. In different traveling problems with the tourist spot-constraints, Beed et al. (2020) combined BA with PSO to optimize tour routes by visitors. They used multi-objectives function to meet the minimum travel cost, distance and time in terms of deviation between tourist spots, and maximized the number of tourist spots to be visited. The combination of these two algorithms shows an improvement to the fitness values by 0.23874 as compared to BA alone which is 0.24963. Yet, this model unable to consider if there are possibilities for visitors to car-pooling and find the least routes in order to maximizes the profit obtained by the traveling agents. Other than that, Long et al. (2014) planned urban road networks using the BA to minimize travel time and vehicle emission cost. Their study made a comparison between a BA solution model with GA solution model and found that BA provides a better solution with standard deviation is 5.06 than GA with standard deviation is 11.36. In addition, the researchers emphasized that the BA could be explored for more factors they suggested to be taken into consideration for the tradeoff between traffic efficiency and traffic-induced environmental pollution when implementing a turning restriction strategy as an effort to further improve their design.

Alzaqebah et al. (2018) optimized the overall transportation costs by minimizing the overall traveling distance for a number of vehicles when serving a set of customers. In their study, they compare the results with GA, TS, SA, PSO, Dantzig-Wolfe algorithm, local search algorithm, and linear programming. Alzaqebah et al. (2018) optimized the overall transportation costs by minimizing the overall traveling distance for a number of vehicles when serving a set of customers. To test the capability of BA in transportation costs problem solving, they compare their results with previous work with similar problems solved, with GA, TS, SA, PSO, Dantzig-Wolfe algorithm, local search algorithm, and linear programming. The solution model with BA shows better performance as compared to previous work by minimizing the travel distance for each car traveled by 75.36 m to 226.43 m than the with the best-known solution model by 82.29 m to 442.86 m. Taking into account the multiple-objective functions to ensure the effectiveness of land use transportation interaction over time to serve the citizens, Szeto et al. (2015) used the BA to examine the sustainability of road network design by considering the changing demands over years and requirement to upgrade the existing network. In addition, their study also took into account dimensions of sustainability that included social, economic, and environmental importance of the area through tradeoff analysis. The solution model of BA in their study provides a better solution in terms of sustainability in urban area and improved the common practice of road network planning and designs in Hong Kong.

During the time this review was conducted, accurate, up-to date and reliable information about the uses of BA in forest road transportation was not available. Yet, all work discussed above explained that the BA has the potential to be applied in forestry contexts since this algorithm has generally shown better performance than other algorithms in terms of optimization speed and the performance of their results by using the BA solution model. In addition, the BA is capable in both global and local search to fulfill the criterion for employment in the field of forestry.

4.1. Potential uses of bees algorithm in Malaysian

We were going to propose a BA for solving a fixed and variable costs of forest transportation planning and design in hill Malaysian forest. The model, taken from the literatures, involves a multi-constraint viz. restricted to elevation, slope, river crossing, multi-facility viz. forest road facility differs according to different constraints and uses of more than one timber extraction machine (i.e. CT and LF)- combination of more than one machinery are expected to reduce cost, or might incur high cost in planning, design and maintenance. The speciality of BA and the potential to use in forest transportation planning is due to its accuracy and less processing time in problem-solving (Pham et al. 2005) that bring the idea to be applied in hill forest area in Malaysia.

In Malaysian forest transportation planning, planners always confronted with multi-constraints. Increment and decrement of operational cost can be seen when multi-constraints minimized at the same time. Among constraint always to be considered in forest transportation planning are; elevation below 1000 m (Thang 1987; Jusoff and Taha 2008) or within production forest, slope less than 20% if CT (Jusoff and Taha 2008) is to be used, and ≤40% if LF is chosen (Norizah et al. 2014). Boundary of river depending on the slope of river itself following Equation [7.6 + (0.6 x % slope)] (FDPM 1999; Jusoff and Taha 2008) are also taken into consideration as constraints for least cost forest transportation planning. Location of existing road is used in order to minimize the area opening which later helps to reduce cost of forest road construction and maintenance (Jusoff and Taha 2008; Norizah et al. 2014) such done by Norizah and Chung (2014) in previous work. Selection of machineries through analysis tradeoff; CT or LF also determines the number of constraints; in order to prepare the forest road that suitable to be used by those machineries in a harvest area. Indeed, these constraints led to the use of RIL concept encompassing road construction improvement, directional felling, good timber extraction practices and choices of suitable harvesting machineries.

Although previous research discussed above focused on road network problem in urban areas with BA, but the concepts of solutions they tried to elucidate were similar with forest transportation problem in Malaysian forest. For example, the extraction of log from stump site to landing can be represented with the concepts of serving the communities to reach their targeted location within a minimum travel time and finding the shortest routes (Long et al. 2014; Iqbal et al. 2015; Ng et al. 2017; Alzaqebah et al. 2018; Chen and Zhou 2018; George and Binu 2018; Santosh and Suresh 2019; Beed et al. 2020; Karaoglan et al. 2020). By shortening the traveling time, the cost for timber extraction can be reduced, thus the operational efficiency can be increased. Such stressed in a study by Santosh and Suresh (2019), the operational efficiency is about reducing the time, distance and costs. It is worth highlighting the fact that transportation operation accounts for a significant part of wood procurement cost involving fixed and variable costs. It is therefore necessary to reduce the overall transportation cost through enhanced or well-organized timber extraction planning and transportation networks.

The wise use of existing road in timber harvest area are quite similar with upgrading the existing road network to reduce construction costs and to overcome the tradeoff dimension of sustainability objectives by Szeto et al. (2015) and analysis tradeoff between two timber transport activities; CT and LF have the similar concept with analysis tradeoff between traffic efficiency and vehicle passes with less impact to environmental pollution studied by Long et al. (2014). In a review conducted by Audy et al. (2012), they found the analysis tradeoff for the sustainability dimensions; environment, economic and social at the strategic transportation planning level may benefit the stakeholders and communities. By confirming the types of machineries to deliver timber from the harvest areas to the final destinations, the next concerns related to road specifications and infrastructures can be determined.

5. Summary overview

The purpose of this review paper is to provide a comprehensive overview of the application of algorithms in solving the forest transportation planning problem. The application of algorithms in the forest transportation planning will help decision makers to optimize solutions to their multi-problems, as well as minimizing or maximizing the objective functions, while taking several constraints into consideration. A number of algorithm techniques have been found to be feasible for the purpose of forest transportation planning. Based on the related literature reviewed, most studies have the same goal, which is reducing the overall operational cost of timber harvesting. This review paper attempts to ascertain the criteria that enable the BA to be a workable approach with added value; the algorithm can then be applied to solve the forest transportation planning problem, particularly in Malaysia. Several previous studies such as those of Pham et al. (2006), Pham et al. (2007), Alfi et al. (2011), Abdel Hamid et al. (2013) Luo et al. (2014) and Farajvand et al. (2018), suggest that the BA is capable of yielding accurate results within a short period of time (number of iterations decreased); this can be a good reason to test this algorithm in solving the forest transportation planning problem. The idea comes from the experiment of solving the least-cost transportation problem conducted by Alzaqebah et al. (2018) and Szeto et al. (2015), who attempted to plan a transportation network by minimizing the cost while taking into account several constraints. Eventually, the author of this review hopes to share the general concepts of how BA can be a useful algorithm in solving the forest transportation planning problem, especially in Malaysia.

Disclosure statement

No potential conflict of interest was reported by the author(s).