Optimal Dynamic Routing for 2 Forklifts in Narrow-Aisle Racking Warehouse

Use your smartphone to scan this QR code and download this article ABSTRACT Determining storage location and planning path are the twomost important components in warehouse management. Simultaneous resolution of these problems not only reduces the storage and retrieval time but also avoid the loss of goods. The article offers a scenario of a practical cold warehouse systemwith narrow aisle racking, where space optimization and time scheduling are always top priority. There are 2 forklift were considered to work parallel in system so aisle dispute is considered to minimize safety risks in warehouse. Two algorithms used to optimize the path were introduced in the paper, which is the closest open location – COL and A star algorithm. The COL helps to determine the most appropriate storage location according to the user's requirements, including type of goods to be exported or imported, finding storage location of the nearest empty cell by refer the weight of the road and obstacle were might happen by two forklift truck in the system. The result of this algorithm are determined Input and Output point of each forklift path. The coordinate index of these two points are returned as input to the A star algorithm to determine the shortest path for the forklift. With the A star algorithm, a clear path will be sought, including the comparison of clashes between vehicles in the system, preferring the shortest path for moving between two points. The travel route results are exported for goods execution devices. The system is simulated byMATLAB combined with V-Rep software for an intuitive interface and fully illustrates each task of each vehicle from time to time. Some traditional or single algorithms with the same assumptions about the system were also simulated and compared to see the effectiveness of the combination of two COL and A star algorithms in a narrow aisle racking system.


INTRODUCTION
Determining storage location -localization is under-2 stood as the process of selecting the optimal storage 3 location among different position, so that the travel 4 time is minimization, thus saving the total operating 5 costs of the warehouse. In addition, each storage loca- 6 tion should be closely managed based on information 7 such as type of goods, stored time, coordinates. Plan-8 ning path is defined as a process for selecting the most 9 optimal path from all the solutions. The optimal path 10 is determined based on two factors: the distance from  Assumption made Layout design 26 The system is built based on some special characteris-27 tics of cold store for preservation of aquatic products 28 but we can adjust it to suit different types of storage.

34
• Goods are organized into the pallet. Each pallet is 35 a SKU. This is the smallest item in system. Pallet is 36 placed on single pallet racking and other picker can 37 reach all items in the rack regardless of rack's height. 38 • Pick out time is undefined for all SKUs in system.

39
For frozen shrimp products, the requirement in stor-40 age process is if goods were come first, it will be sorted 41 in pallet racking first (First Come First Served -FCFS) 42 and travel distance for each moving cycle is the short-43 est to prevent damage under wrong temperature. In

82
Evaluation function: • Operating cost function, g(n) -Actual operating 84 cost having been already traversed.

85
• Heuristic function, h(n) -Information used to find 86 the promising node to traverse, the heuristic function 87 must be admissible.

88
Each storage location in warehouse is represented by 89 a node, it will be used as an object of the algorithm in 90 this section. Some notations is given below: The g(n best ) in this flow chart represents the exact 98 travel distance of the path from the starting point 99 to any vertex n best -which is defined as a shortest 100 node in each step of the loop and h(n best ) represents 101 the heuristic estimated distance from vertex n best to 102 the selected storage location x. h(n) value is calcu-103 lated using the Euclidean distance formula. Each time 104 through the main loop, it examines the vertex n that 105 has the lowest (1) with each: One more node in the shortest path is found. The 107 main loop repeat until latest node is determined -108 which represent selected storage location.

109
The auto-localization algorithm base on A-star ap-110 proach is clear. It is easy to implement and allows very 111 fast route computations since this method only cares 112 about the start and end of each row and ignore the 113 time dependent between forklifts. How-ever, when 114 system was performed by 2 forklifts, various draw-115 backs are caused by deadlock and traffic jam have a 116 deteriorating effect on the system performance (see 117 Figure 3).

118
To deal with the problems of the model given in previ-119 ous Section, a different approach that computes short-120 est (traveling time) and conflict-free routes simulta-121 neously is propose which time-dependent between 122 vehicles is considered 5,7,10 . The idea of the algorithm is that find a conflict-free 138 shortest-time route in the case there is collision po-139 tential in the aisle (see Figure 4). According to the 140 approach, after the shortest path to the storage loca-141 tion is found by the A* algorithm, a time-dependent 142 histogram is established. Based on distance and veloc-143 ity data, the position of each vehicle at each time on 144 the map is determined and then a free-conflict path is 145 formed by using the waiting time for the vehicle (see 146 Figure 5).  the ware-house layout instead of using complex algo-173 rithms to find an optimal path for single pick aisle as 174 the article (see Figure 9). For this method, the algo-175 rithm to localization is same with dynamic routing ap-176 proach.

177
The comparison result shown that travel time of opti-178 mal dynamic routing is approximately 7.33% less than 179 double aisle approach (shown in Figure 10)