Example: Supply Chain
Today’s supply chains contain inefficiencies, which manifest themselves in both sub-optimal performance in specific functions, and also in the holistic flow of items through the entire chain. For example, over 30 million containers go through ports in the US alone and navigating each of these containers through the system efficiently is a complex problem with many sub-components; offloading from ship to shore, moving containers to the stacks, and then to transport gateways. There are complications with regards to customs paperwork, chassis availability, union regulations, and clean air initiatives. Each of these sub-processes demand attention to optimize, which is challenged by the fact that an improvement in one area frequently has a conflicting impact elsewhere. In addition, each process typically involves many organizations, meaning that central control of a solution is nearly impossible to achieve.
Swarm allows each participant to model their own operation, and optimise it independently. We also allow participants to collaborate, and share a dynamic model that optimises the entire process from end-end
Mixing raw materials together, or intermediate products, is a common step in Food Production. This is typically done for reasons of flavor, cost, to remove contaminants, add benefits (e.g. protein) and to meet regulatory requirements. At a relatively low level the number of combinations becomes astronomical, overwhelming a modern cloud-based analytic system. Many blends are also non-linear – the viscosity of two liquids is not the average – and require individual formulas or lab-based heuristics. These environments also suffer from common disruptions like corroded pipes, or shipping delays. While optimization software exists today, most organizations find the solutions lacking, as the overall complexity and supply chain integration issues contribute to sub-optimal blends, higher costs and poor margins.
SWARM can utilise a range of algorithms in scenarios where analytics fail because of the sheer number of options. By combining this with reinforcement learning, we can then optimise across a specific time period - such as a financial quarter, or a growing season