Blog
Select Page

How Transmetrics Empowers Logistics Optimization with IBM Watson

01.06.2021

Global trade has reached peak performance, with ever-increasing volumes of goods transported around the globe at astounding speed, never seen before transparency, and mind-blowing accuracy. A 1921 postman would probably think an alien civilization took over when seeing a modern distribution hub in 2021. Modern hubs are full of robotics and tools the postman couldn’t even begin to understand such as autonomous vehicles, smart baggage belts, and an array of sensors to keep them in check. Due to the rising standards set by us, the customers, the business of getting shipments from A to B has changed dramatically. We require goods to be at our doorstep with ‘priority mail’ or even ‘same-day delivery’ without a fuss. 

You can imagine, therefore, that the pressure on the people behind the scenes has started to rise considerably as well. In order to cope with high demand and provide customers with fast deliveries, some warehouse employees have to work 12-hour shifts with limited breaks in order to meet the daily rate of processed packages, a KPI used by many companies including Amazon. In the meantime, there are also challenges for another extremely important role in the delivery chain – truck drivers. Some truck drivers report a high rate of personal and health issues due to the always on-the-road lifestyle. Such rough conditions also negatively influence the inflow of new workforce with some European logistics firms expecting the number of unfilled driver roles to leap to 17% over the course of 2021. All of these raise safety and ethical concerns for the people working in the industry while pushing companies to look for alternative solutions and involve technology to improve the working conditions and optimize the delivery chain.

Given these circumstances, the problem of logistics optimization became a major challenge for the industry. When we say “logistics optimization”, what we mean by that is the structural, intentional usage of state-of-the-art IT systems to enhance the performance of activities related to the logistics value chain. The amount of variables that have to be considered for this optimization is difficult to comprehend even for the most experienced logistics professionals. For example, not every shipment goes straight from point A to point B. In fact, the outstanding majority of goods have to travel distances of thousands of kilometers or more, go through multiple hubs, ports, customs and the amount of potential pathways increases exponentially with every possible stopping point.

On top of that, companies have to account for the travel time, transport mode, staff work hours, choosing between express vs. regular shipping options while making sure that there is just enough capacity to handle all of this. It is already a tremendous task to analyze such operations retrospectively or in real-time. Demand forecasting and further optimization bring yet another level of complexity, adding the historical data and external factors into the calculation.

Considering the above, identifying the most optimal path to transport just one shipment from point A to point B becomes an impossible task for a human brain. Data Analytics and Artificial Intelligence provide logistics companies with the ability to process these large data sets that include all the relevant variables and drive conclusions that enable decision-making on a level that was previously impossible. Currently, only 12% of logistics companies use AI in operations, but it is expected to grow to 60% by 2025. 

When Transmetrics founders started the business in 2013, the industry has been looking at AI as something that can only exist in science fiction. Fast forward 8 years and Artificial Intelligence became one of the hottest topics in the sector with dozens of proven use cases and tangible benefits experienced by the industry players. Since our team has been building the AI in logistics expertise from day one, it shaped Transmetrics into one of the frontrunners of, perhaps, the most exciting industrial shifts of the past decades.

However, our mission is still the same as it was at the beginning – we want to empower our clients to be at the forefront of this revolutionary AI shift. In order to do so, our team has spent years developing the Transmetrics platform and fine-tuning state-of-the-art algorithms that derive useful insights by applying data evaluations, predictive analytics, forecasting, and optimization models. All these steps are supporting one important goal – helping logistics organizations make the optimal data-driven decisions ahead of the demand curve and pushing their service quality to the next level.

However, such optimization might be extremely complex and in order to provide our clients and partners with the most insightful and accurate estimations, Transmetrics has partnered with IBM in order to leverage the power of IBM Watson in order to solve some of the most complex network planning problems that the logistics industry has to offer. 

Artificial Intelligence: the IBM perspective

According to Gartner, in 2021 AI augmentation will generate 2,9 trillion US dollars in business value and recover 6,2 billion hours of worker productivity. But how do you really take advantage of AI-based technology and applications?

IBM Watson is the answer. Watson is the AI technology for business, helping organizations to make more accurate predictions, automate processes, interact with users and customers, and augment expertise. Watson has evolved from an IBM Research project to experimentation, to a scaled, open set of products that run anywhere. With more than 40,000 client engagements, Watson is being applied by leading global brands across a variety of industries to transform how people work. IBM is focused on AI for business and helping customers reclaim precious time through automation, generate critical insights through natural language processing, and foster trust in outcomes derived from AI. To give businesses, data scientists, and developers the capabilities they need to scale AI, IBM is continually enhancing and expanding the pool of innovations offered in IBM Watson from IBM Research. 

Recently, the global research firm Gartner has positioned IBM as a Leader in two Gartner Magic Quadrant reports – the 2021 Magic Quadrant for Cloud AI Developer Services and the 2021 Magic Quadrant for Data Science and Machine Learning Platforms. These new recognitions from Gartner underscore the value IBM is bringing to businesses through strategy and a strong innovation pipeline between IBM Research and IBM Watson. All that to benefit the business in various industries, including transportation and logistics.

Watson Studio is the tool in the IBM portfolio that empowers developers to operationalize AI anywhere. It helps to build and scale AI with trust and transparency by automating AI lifecycle management. One of the many benefits organizations can gain is the ability to optimize business decisions, develop and deploy optimization models quickly and determine the best course of action to improve planning and scheduling outcomes. For example, a high-performance mathematical programming solver for linear programming, mixed-integer programming, and quadratic programming, CPLEX, enables decision optimization for improving efficiency, reducing costs, and increasing profitability.

With IBM organizations across various industries have access to a robust ecosystem of partners that helps them seize the AI opportunity. IBM business partners, like Transmetrics, leverage IBM Watson to deliver state-of-the-art solutions for clients and help them easily adopt AI for their businesses. This is truly a game-changer for many organizations that can take advantage of AI-based technology in a quick and efficient way, transforming their operations and how they work.

How does IBM enhance Transmetrics optimization capabilities?

Partnering with IBM brings many benefits and improves Transmetrics operations in several aspects. However, all of these aspects serve one major goal –  enhance the value of the Transmetrics optimization for our clients.

Cloud First

IBM Watson studio allows our team to access world-class experiment tracking, collaboration tools, API building, and enables fast, easy, and reliable model deployment. This enables us to transition more effortlessly into a data-driven organization, where everybody can easily use the cutting-edge developments of the Transmetrics research and data engineering teams and put it in the hands of internal and external customers. By using the IBM cloud capabilities, our teams can easily explore ideas while worrying less about computational resources, experiment tracking, and assessing the performance.

The studio interface is very convenient, integrating familiar interfaces such as Jupyter notebooks, and facilities for quick dashboarding and model deployment. Using its capabilities, the data science and engineering teams can easily share prototypes and Proofs of Concept internally with the product and project teams. We can play around with ideas together more easily allowing faster and more pleasant feedback loops, ultimately decreasing time-to-market.  

Assessing the complexity of the problem

Transmetrics’ main use case of IBM Watson is related to decision optimization since we leverage mixed-integer and mixed-integer quadratic programming to achieve that.

However, before we dive deeper into the exact applications, we have to explain the specifics of the problem our clients face which is a variation of the Stochastic Multicommodity Fixed-charge Service Transport Network Design problem. Or, in simpler terms – a Linehaul Planning problem (LPP)

Below is the list of given variables that Transmetrics takes into account to solve the LPP by focusing on one major objective – minimizing fixed and operational costs for a given service level which translates into the probability of delivering a particular package on time:

  • A set of commodities each with an origin, destination, time of receiving, delivery deadline, and distribution of demand quantity.
  • A set of vehicles with unit usage costs and capacities.
  • A set of potential contracts with subcontractors, which have fixed prices and predetermined routes. Once a contract has been purchased, it could be used for the whole period.
  • Cross-docking capacities at a given warehouse/hub, potentially different for different hours of the day that can be expanded on-demand for an additional cost.
  • Possibility to acquire additional trips flexibly on the spot market for an additional cost.

 

A world-class prescriptive analytics engine

 

Figure 1, a visual representation of our software running a network analysis on different hubs in Europe.

As you can see, the sheer number of different possibilities in such variables is something that a human brain cannot process, even more so use them for optimization. This is where we use Transmetrics proprietary AI algorithms and our expertise in logistics planning to generate suggestions for our clients. On top of that we use the layer of IBM’s CPLEX tools, which provides us with the most optimal and accurate optimization results within the following applications:

  1. Finding high-quality solutions to large Network planning problems

For example, we’ve been able to find high-quality solutions to problems 90x the size of the programs we were able to tackle previously. While open-source solver finds none or significantly inferior solutions for a network planning problem with 12 sites with a 7-day time horizon, CPLEX finds near-optimal solutions with more than 100 sites in minutes.

  1. Modifying results based on a feedback loop

Solving modified problems on the fly, enabling on-demand planning updates as new information arrives or if the planner’s preferences differ from the proposed solution. For example, in a network planning setting, a local planner might not like the proposed dispatch for certain days. We can then give them an interface to make corrections, and show a corrected plan for the whole network and demonstrate how the global solution would be impacted and if any unintended consequences would occur.

  1. Migrating towards stochastic and robust programming

Stochastic optimization methods provide more robust and reliable planning solutions by taking uncertainty into account. Unfortunately, their complexity grows immensely with the sample size and number of periods ahead that we need to look at. Thus, if we want to find near-optimal solutions for a given time period, it’s imperative to use specialized algorithms to do so. With CPLEX, Transmetrics can access the best-in-class ability to unlock solutions to such problems – for example, via creating sophisticated decomposition with callbacks and advanced solving strategies. In particular, for a use case with 50 scenarios, the open-source solvers find a high-quality solution (within 2 percent of the optimum) within an hour in 5% of the cases, compared to CPLEX, which finds such solutions in 95% of the time. This translates to about a 2% improvement in the operational costs for this leg of the operations.

The Next Big Step

With IBM Watson empowering the Transmetrics optimization efforts, our team can ensure that clients receive the most accurate and up-to-date planning and optimization suggestions. However, logistics planners are always in control: they decide to accept, modify or even reject the suggestions that the platform provides. This ensures we always have checks and balances in place while simultaneously it provides the opportunity for the platform to constantly learn from new inputs and enhances the working conditions for the users of our solutions instead of replacing their jobs. Based on our experience, combining the computational power of AI with the experience of a logistics professional always brings the most significant benefits to our clients.

For example, one of our clients, DPD Group, has significantly improved the capacity utilization in their network and optimized their linehaul planning. By leveraging AI and proprietary forecasting and optimization algorithms developed by Transmetrics, DPD has reduced its hub-to-hub costs by 25%, with a 14% increase in fleet utilization. Total cost reduction is estimated at an astonishing 7-9%. In short, DPD has increased its service levels while improving a bottom line, the holy grail in transportation.

While AI is still a relatively new technology in the logistics industry, the potential positive impact of it cannot be overstated. According to the report from Meticulous Research, Artificial Intelligence in the Logistics market will be worth $21.8 billion by 2027. It is not anymore a matter of if, but a matter of when your company will introduce AI technology into your operations. The sooner it happens, the more competitive, profitable, and tech-driven your logistics business will become.