We can help you utilize your data in the strategic, tactical and operative planning stages. Based on our vast experience with the methodologies of both data analysis (descriptive, predictive and prescriptive) and Operations Research, we will develop individual solutions for diverse problems in the fields of production and logistics.
Increase your process efficiency with our expertise in the field of artificial intelligence. Neither a large amount of data nor a precise model description are necessary in advance. Together, we will characterize and quantify what defines a »good« process for you individually, and which influencing factors may play an important role. On this basis, reinforcement learning methods can be used to learn a strategy that is optimal for your process independently.
In cycles, consisting of 3 steps, the agent learns over time which decisions lead to a good process and which do not. In the first step, the current state of the environment is evaluated. Based on this, the agent chooses, in the second step, an action (reaction) from a predefined action space. At this point, the right balance between using existing knowledge and generating new knowledge is essential. In the third step, the selected action is quantitatively evaluated based on the system behaviour. In doing so, the agent expands its knowledge about the behaviour of your system cycle by cycle which leads directly to an improvement of your process.
Customized Loading Space Optimization
When it comes to non-standardized goods, it is hard to determine during the planning stage which order combinations require how much loading space. The course of action in these cases is usually to rely on estimations and to leave some buffer space, which leads to empty space in the vehicles and an unnecessarily high number of trips. We develop solution algorithms that optimize load planning for non-standardized goods and individual loading methods. Based on our wide experience and continuously growing TrueLoad toolbox, we will create your personal planning tool according to your specific requirements, so that you will know even in the planning stage what will fit into your vehicle and how to manage the loading.
Interaction of Vehicle Routing and Production Planning
In just-in-time production, supply and production dates are coordinated. Thus, production, loading and delivery are closely linked. Despite this fact, however, the planning of each of the three steps is usually done separately and without direct synchronization. In many cases, route planning, which is carried out first, indirectly determines what needs to be manufactured on which day. However, this planning step often only considers vague capacity potentials and ignores other framework conditions that can guarantee a smooth production process (e.g., product mix). As a result, there are frequent disruptions during the production process, which, in turn, affect loading and delivery. Within the OptiCap research project, we combined the three planning levels for production, loading and vehicle routing in order to better harmonize production and shipping, thus achieving optimal use of capacities.
Learning from the Past
Planning tasks in the fields of production, logistics and infrastructure management are often scheduled on the basis of data that is unchanged over a very long period of time. For example, processing times that have been recorded only once are used for product types and process steps even though they are subject to significant variations. As a consequence, there often are deviations from the plan in everyday operations. The Fraunhofer IVI works on the task of feeding back experience gained into the planning process. By continuously comparing planned processes with their actual realization, as well as by the analysis of correlations and the integration of that knowledge into the data basis, robust and realistic planning is made possible.
Vehicle Routing System with Time-of-Day-Based Travel Times
Everyone knows that if you have to drive across town in rush hour, you should schedule for more time. This fact is a special problem for shipping service providers and forwarders, because varying travel times are often ignored in vehicle routing. The result are delays, and buffer time needs to be provided in order to comply with fixed delivery dates specified in a contract.
Within the TOTARI research project, we tackled this challenge. Most delivery vehicles are equipped with a GPS device that records driving data such as positions and speeds. This data is analyzed and attributed to individual route segments in order to determine a time-of-day-based speed profile. The vehicle routing system then calculates different speeds depending on time of day and route segment.
Fraunhofer Institute for Transportation and Infrastructure Systems IVI