How can we solve distribution problems?
«Supply chain management practices are becoming increasingly removed from what we understand under the term,» according to Professor Wolfgang Stölzle, director of the logistics management faculty at the University of St Gallen (Switzerland). He presented his assessment of the segment at the opening of his institution’s so-called logistics get-together.
Around 120 guests recently attended the St Gallen logistics event, which has become a minor tradition in the meantime. They did not have long to wait for evidence of the accuracy of Stölzle’s assessment of supply chain management, as it came as early as the second presentation, which bore the curious title of «Transport planning using ant logistics». This was a collaborative effort between Peter Geiger, head of national transport at Migros, Switzerland’s leading retailer, and Dr Ruedi Hug, managing director of the software and consultancy firm Cantaluppi & Hug.
The two companies have been collaborating in various fields for more than two decades. In 2002, with the commissioning of Migros’ new Suhr distribution centre in the canton of Aargau (Switzerland), initial attempts were made to address the national distribution problems using ant logistics. In 2011 and 2012 the partners again turned to the hymenoptera principle to optimise planning. Their project won a Swiss logistics award from GS1 Schweiz in 2012. Migros’ logistics activities are impressive. 400 rail wagons plus 500 to 1,000 trucks are on the move every day. Over and above this, about 2,000 transport requests are received and 600 sites (supermarkets, specialist stores and restaurants) are supplied on a daily basis. The challenges associated with scheduling all this transport are difficult.
«Our system is extremely complex»
These include quality requirements (temperature monitoring), branch restrictions (ramps, journey times, height limits), distribution centre constraints (loading times), seasonal fluctuations in delivery volumes (Easter, Christmas, special offers), vehicles (type, payload, refrigerated or unrefrigerated), changing delivery schedules (for example on account of public holidays), the road situation (closures, peak traffic periods) and extreme time pressure.
«Our system has become extremely complex. Every day, we try to reorganise things and seek to automate our processes,» Geiger explained. His vision contrasts the status quo, aiming to replace static transport planning (fixed trip frameworks combined with selective optimisation measures) with more dynamic solutions (based on actual values).
Frequent violation of restrictions will have to give way to complying with predefined constraints. In addition, the enormous amount of time required (scheduler’s priority with order management) should be replaced by short calculation times (scheduler’s priority with quality assurance).
Problem-solving ant algorithms
On top of that, the company has to deal with a great deal of potential combinations, Hug underscored, citing two impressive examples. Whilst it takes around 1 second to calculate the needs of 40 customers with 1,000,000,000,000 possible combinations, it takes about 1,000 years to calculate the requirements of 70 customers with 1,281,600,000,000,000,000,000 possible combinations.
«There isn’t an optimal solution, nor any evidence that one has been found,» conceded Hug, elaborating that «we first hit on the ideas of ants at the end of the 1990s.» Ant algorithms are a scientific way of solving problems. The colony-building insects are able to optimise accessible routes to food sources from their nest, finding the shortest route. This is now simulated by computer.
The animals, which are almost blind, are unable to see where they are going, so they work together (colony mentality), leaving pheromone trails along the way. These scents, which evaporate over time, assist communication – along with the use of antennae. The more intense the smell, the better the trail, commented the 61-year-old. When translated into route planning, this means that determining the ideal distribution plan is an iterative process.
In this way, many thousand distribution plans have been developed. Successful routes (with more pheromone scent) have a greater chance of being reselected. In the optimum transport plan, all routes are awarded a success factor (depending on the pheromone intensity). After a predetermined number of iterations, or when the time target has been achieved, the process is terminated and the most favourable plan adopted. This represents the chosen solution. The success factors of ant logistics, which according to Hug can also be used for other purposes such as container management at ports, are obvious. It is a high-speed method of meta-heuristics, easy to implement, depending on conditions, and follows the principle of memory (colony mentality). «It’s still in its infancy, however,» Hug qualified.
Schedulers recognise scented trail
Other elements of success include the validation of the solution’s plausibility by the scheduler, «who knows the best route and thus carries the scented trail in his head. But there shouldn’t be any silly solutions; this aspect is the target of our work,» clarified Hug. Ant logistics enable comprehensive planning: transport-controlled order picking, integrated driver support and real-time information to the consignee.
By using ant logistics, the scheduler is transformed from a planner into a transport controller, is how Geiger described the link. Now more than ever, he has to bear three aspects in mind: customer orientation and optimum fulfilment of the transport request, the lowest possible transport costs (a high degree of vehicle utilisation, minimal route kilometres) and time savings for the benefit of maximum quality.
48-year-old Geiger estimates the optimisation potential of today’s transport costs, which Hug described as colossal due to the large number of routes, at up to 10%. Implementation is still on-going. So far, the areas of transport simulation/quotation and planning seasonal business/promotional distribution are already being used productively, whilst the issue of dynamic daily route planning still lags behind. Migros is now at the stage of using the optimisation tool parallel to daily planning, but still far from achieving its goal. In terms of transport simulation/quotations, Migros also does not have everything under control, conceded Hug and Geiger in unison. Their conclusion was that «we’re in the process of solving an old problem – thanks to a completely new approach.» Mindful of the challenges they still have to face, Stölzle hoped they will both persevere.
Log HSG has an excellent network
The logistics management faculty at the University of St Gallen (Log HSG) maintains a large target group network, as its director, Professor Wolfgang Stölzle, emphasised at his institute’s 7th St Gallen logistics get-together. In addition to a core team of 15 people, which includes nine scientific staff (as well as 14 students and interns in varying combinations), the network consists of the executive committee and an advisory council, alumni of the executive MBA and DLM programmes, project and scientific partners and former employees (postgraduates and interns).
Projects and seminars
Log HSG’s current initiatives include focusing on integrated supply chain management (consistent management up to the raw material source), supply chain differentiation and performance-based contracting (innovative structuring of business relationships), as well as seminars on integrated inventory management (ways to optimise working capital) and on integrated transport management (ways to optimise the (eco) efficiency of road haulage).