Intelligent logistics
"a symphony of software, machine learning, computer algorithms, and people"
[Published 2.11.2017]
Logistics leaders are beginning to capitalize on the immense amounts of data generated by their cyber-physical systems. With artificial intelligence they’re finding greater efficiencies, higher performance, more sophisticated network orchestration, and more accurate demand forecasts.
Logistics is an enormous industry that has wrapped a complex network around the planet. Revenues are estimated at $8 trillion and are expected to grow to $15.5 trillion by 2024. IDC forecasts global spending on artificial intelligence will grow to $47B by 2020 - a %55 CAGR. Logistics companies are now seeing the value of AI as it converges with the industrial Internet of Things.
Like many parts of the supply chain, logistics are experiencing digital transformation. Connecting up the logistics value chain yields more data, more interoperability, and greater control. Now, it’s being amplified by learning systems - assemblies of deep learning algorithms - that sift through the data, build and manage high-resolution models, fine tune operational performance, and predict changes across the value chain. For logistics operators, these changes enable continuous sensing and learning, and the ability to be agile and adaptive in an accelerating world.
The German logistics company DHL sees this as a $1.9 trillion opportunity. Citing pressures from energy, security, and technology, their Logistics Trends Radar declares, “the impact of data-driven and autonomous supply chains provides an opportunity for previously unimaginable levels of optimization in manufacturing, logistics, warehousing and last mile delivery that could become a reality in less than half a decade.” Logistics is entering the 4th Industrial Revolution.
"...the chance to lower energy and water usage, pollution and emissions, and the adjacencies of industrialization."
Supply & demand
From inventory to delivery, the industry has continued to digitize and automate. Amazon now deploys over 45,000 robotic pickers in its warehouses. Learning systems visually inspect new packages, monitor shelving, allocate packing for each item, and orchestrate the movement of inventory from the warehouse onto shipping vehicles. This convergence of data, robotics, and machine intelligence has become a template for cyber-physical systems. “We like to think of it as a symphony of software, machine learning, computer algorithms, and people,” says Amazon spokeswoman Kelly Cheeseman.
Managing supply and demand can be highly complex and missteps can mean missed opportunities and wasted goods, especially for perishable items. Shelf Engine is using learning systems to forecast how much food to order based on historic demand patterns. They helped increase profits for retailer Molly’s by 9%. Multinational grocer Tesco saved $140 million by feeding weather data into their predictive analytics engine to forecast demand for weather-related foods like ice cream. They uses this data to adjust inventory and orders for each store, orchestrating their dynamic supply network around diverse geographical demand.
From inventory to orchestration
Logistics companies must coordinate large networks of vendors, partners, sub-contractors, and vehicles. These exist within larger contexts like regional demand, global economics, geopolitics, and climate. As learning systems build stronger models and integrate information from the world, their ability to act as network orchestrators is growing. ClearMetal tackles this complexity using predictive logistics to manage shipping containers. To reduce risk between supply and demand, they track containers, evaluate carriers, and integrate 3rd party signals about weather, economics, and port conditions, modeling it all with machine learning.
Under the hood
As more components become connected and sensing, more data can be fed into learning systems. Rolls Royce’s Trent XWB jet engine has thousands of sensors producing terabytes of data on long-haul flights.8 Using Microsoft’s Azure Stream Analytics and Machine Intelligence libraries they optimize fuel efficiency and anticipate maintenance, saving costs and minimizing downtime. GE Transportation does the same with its locomotives while trying to instrument a U.S. rail system that carries 5 million tons of freight every day. Daimler and Navistar have each piloted autonomous long-haul trucks in connected platoons - Anheuser-Busch InBev delivered a truckload of beer autonomously (with the help of Uber’s Otto driverless unit). AB InBev says autonomous trucking could save $50 million annually in the US.
Mumbai’s ABB Ship Management shows the roadmap for many heavy industries. They’re building networks on and between their ships, and services to connect and automate operations. This will let them use remote controls and then launch fully autonomous systems. ABB shows how instrumenting physical assets to capture high-resolution data lays the foundation for deep analytics, predictive models, and self-management.
Implications
In their work on the 4th Industrial Revolution, the World Economic Forum has declared that “we stand on the brink of a technological revolution that will fundamentally alter the way we live, work, and relate to one another.” Leaders in logistics have been steadily laying this foundation and are now building learning systems to make sense of all the data. Startups are trying to unbundle logistics, allowing large operators some flexibility and the opportunity to divest while also threatening their control. This is driving more competition and M&A.
The ability to automate, forecast, and orchestrate complex, dynamic ecosystems towards higher orders of performance and efficiency will continue to draw R&D and underwrite quarterly earnings. Google has shown its DeepMind learning systems can reduce energy costs in their data centers by 15%. This illustrates perhaps the most valuable opportunity: the chance to lower energy and water usage, pollution and emissions, and the adjacencies of industrialization.
Roadblocks
There are many challenges to transforming logistics. AI is hard and logistics is a large and fragmented industry that coordinates tens-of-thousands of businesses. There are considerable data and interoperability challenges but AI can be trained to translate between them. The industry is risk-averse and slow to adopt, particularly when networking opens up security and vulnerability issues. Furthermore, regulatory responses will be strong as autonomous systems continue to put pressure on human labor. In America, 3.5 million workers drive transport vehicles. As logistics fleets become autonomous, labor crises will develop and push regulators to impose limits.
There are many challenges yet to be addressed but the capabilities offered by AI are too great to be dismissed. This is the next wave of digital transformation that is reaching all industries.
References
Transparency Market Research http://www.transparencymarketresearch.com/pressrelease/logistics-market.htm
IDC http://www.idc.com/getdoc.jsp?containerId=prUS41878616#
DHL Logistics Trend Radar http://www.dhl.com/content/dam/downloads/g0/about_us/logistics_insights/dhl_logistics_trend_radar_2016.pdf
Amazon MIT https://www.technologyreview.com/s/538601/inside-amazons-warehouse-human-robot-symbiosis/
ShelfEngine & Mollys https://static1.squarespace.com/static/5664d42ce4b041edc7625b53/t/588feaa2197aea6ab4068096/1485827289195/Case+Study.pdf
ClearMetal http://thinkapps.com/blog/development/machine-intelligence-clearmetal-interview/
GE Transportation https://sloanreview.mit.edu/case-study/ge-big-bet-on-data-and-analytics/
Daimler & Navistar http://www.computerworld.com/article/3053529/car-tech/a-fleet-of-self-driving-trucks-rumbles-across-europe.html
Anheuser-Busch https://www.bloomberg.com/news/articles/2016-10-25/uber-self-driving-truck-packed-with-budweiser-makes-first-delivery-in-colorado
ABB http://new.abb.com/marine/generations/technical-insight/integrated-operations