On Thursday 26 January, Pipple was invited to speak during the PyData meetup at EyeOn. Within the theme ‘ Demand Forecasting and Logistic Optimization ‘, Wouter Leenen and Lennart van Ham gave a presentation entitled ‘Is there a next step in warehouse logistics optimization?’ For Pipple a well-known topic, which gave an excellent opportunity to share our vision with the public.
“The meetup was a huge success, Lennart and Wouter are awesome presenters, at 21h there were still questions from the audience flying in like there was not tomorrow. Your peer at EyeOn could not believe that people were so enthusiastic at such a late event.” – Gareth Thomas, Organizing Committee Chair & Community Manager PyData
With numerous projects in logistics, Pipple has applied an extensive curriculum of methods and techniques, but the most commonly used are Operations Research (OR) Heuristics. Of course, we would prefer to solve every problem exactly and optimized, but in practice this is not always possible. Heuristics help to simplify a problem definition, and find a good solution in a short time. However, in situations where real-time resolution is needed, the smartest heuristics are often too slow.
In recent years, Reinforcement Learning has made its way within the gaming world. By simulating reality and keeping track of its intelligence in a so-called neural network, RL can make decisions immediately when they are needed. However, an accurate simulation of reality can be very complex, and RL is not suitable for every type of problem. At Pipple, we think we don’t always have to choose between contenders. Instead, we strive for a combination of different techniques. In this way we take advantage of the strengths, and we help to catch the weaknesses of techniques. For example, heuristics can make an RL simulation easier, such as doing route calculations and grouping tasks, and you can choose to apply RL only for the steps in a process that require real-time decision making, such as assigning a task to a freed user. “Not replacing but combining OR-heuristics with innovative Machine Learning technologies.”
One of our key take-aways is ‘Keep an innovative mindset; experiment and experience.’ Partly because of this, it was very nice and very valuable to see speakers from IKEA and EyeOn speak, and to talk to participants about the various topics. In this way we keep our eyes and ears open, and we also learn new things all the time.