Want to start reading immediately? Get a FREE ebook with your print copy when you select the "bundle" option. T+Cs apply.
Artificial Intelligence Applications in the Logistics Industry
Updated: 20th July 2023
Whilst AI has been around for years, it is impossible to avoid the recent media frenzy created by the launch of applications such as ChatGPT or Google’s Bard. So why this sudden burst of interest in the technology, an interest which has driven the share prices of some of the companies involved to stratospheric heights? The answer is that the latest iteration of AI – ‘Artificial Generative Intelligence’ (AGI) is starting to fulfil the original expectations (or hype) for the technology.
AI has often been considered a ‘self-aware’ technology with the ability to think and act autonomously, however, in many cases, this is a misrepresentation of its capabilities. The AI that we have been writing about (forever) is much more limited and refers to systems that are more focused on specific domains or bounded areas of operation, for example managing large numbers of IoT devices, vision systems in warehouses or pricing data related to shipment costs or bookings.
What is AGI?
Tech Target defines AGI as, ‘the representation of generalized human cognitive abilities in software so that, faced with an unfamiliar task, the AGI system could find a solution. The intention of an AGI system is to perform any task that a human being is capable of.’ These systems are trained on colossal datasets of general-purpose information from multiple sources and then essentially directed to get on with it.
It is important to keep in mind that AGI systems have been designed to ‘learn’ through the feedback that occurs in response to any answers they generate. This is because their answers to queries are in many ways, a ‘best guess’, that may be very accurate or total fiction. In much the same way a child learns, AGIs must be corrected when they get something wrong to comprehend what is correct. These incorrect responses that appear to be very plausible are also called ‘hallucinations’ because the system has made something up. This is why any output from an AGI must be qualified by its operators until it is regularly generating reliable answers to queries.
As the underlying hardware and data sources improve, the performance of AGIs will also improve, which should be at an accelerated rate. In the meantime, AGI output should be qualified and checked before use.
The past and present of AI
The earlier versions of AI referred to machines capable of performing tasks that would formerly require human intelligence (such as visual perception, speech recognition, decision-making and language translation). A fundamental component has been the ability of a computer to identify patterns in streams of inputs and learn by association. Through this process, a computer can ‘learn’ to distinguish a dog from a cat by filtering a data bank of thousands of categorised images and responding to human corrections to build an association between the data; Big Data is the ‘fuel’ for AI.
The latest version of AI now includes a combination of specific AI algorithms and advanced hardware such as vision systems, which can be implemented within operations to augment human decision making. Amazon is deploying this technology to identify damaged items within their automated warehouses as they are picked and packed for shipment. In the facilities where these are in operation, throughput has increased and returns of damaged goods have reduced.
Because this is a machine-driven exercise, the data can be analysed to help understand the root cause of the damage to help rectify the problem at the source, which may be poor-quality manufacture, poor-quality packaging or incorrect handling and storage. This is a clear illustration of how the specific application of AI will enhance operational performance in a consistent and reliable manner.
As these technologies progress and mature, they will be increasingly bedded within a mutually supportive ecosystem which operates and improves physical and virtual networks, such as supply chains.
Four applications of AI in logistics
1. Last-mile delivery
One area of logistics which will be increasingly influenced by AI is the operation of last-mile delivery systems. Overall, customers are becoming more demanding. Any company offering a flexible range of delivery options faces an increasingly difficult task in coordinating last-mile flows and this is where the application of AI can dramatically improve delivery services.
Besides optimizing the distribution of shipments, AI can also submit alternatives by crunching customer data; for example, proposing that a customer pick up their consignment from a designated access point, based on geo-location data showing that it will be located on their route home as they commute from work.
By analysing consumer behaviours and location data provided by mobile devices, it is likely that AI will enable companies to become increasingly capable of customizing delivery options for individual customers.
2. Autonomous vehicles
Arguably the most visible manifestation of AI within the e-commerce supply chain is autonomous vehicles. Although significant progress has been made, the likelihood of full autonomy – or ‘driverless trucks’ – is still some way off.
Current programming techniques employed in AI are unable to provide a computer with the ability to infer potential actions in an unfamiliar situation. Without access to any data relating to a similar situation, none of today’s autonomous systems can determine how to respond to extremely low-probability events.
Moreover, conceiving of such events in order to simulate an AI response is challenging in and of itself. As such, the physical safety of autonomous vehicles is incredibly difficult to determine because it is essentially impossible to test exhaustively.
As the last-mile delivery of products constitutes the only visible segment of the supply chain for individual consumers, ‘drones’ have also captured the popular imagination and account for a sizeable proportion of the news coverage relating to AI.
3. Warehouse automation
Warehouse automation has already been significantly impacted by AI. The distribution of products in an Amazon warehouse, for example, is not predetermined by category but uses an organic shelving system where products are arranged by the company’s Warehouse Management System (WMS) using algorithms to optimize placement based on picking routes.
Amazon’s 2012 acquisition of Kiva systems allowed it to optimize fulfilment further, by deploying robots to streamline the picking process; bringing the shelves of goods to the human picker, rather than vice versa. This has enhanced the speed of picking operations within the company’s facilities.
Additional companies have exploited machine learning to ‘train’ autonomous guided vehicles (AGVs) to operate in a mixed warehouse setting, alongside humans.
4. Trade process facilitation
The customs process is highly complicated, involving the entry of masses of data related to entry protocols and procedures for both imports and exports. As such, the vast data sets involved, and the complicated and time-consuming nature of manual entry, make the function appropriate for disruption by AI. The process requires the input of data such as:
- Country of origin
- Description of the goods
- Cost, insurance and freight (CIF)
- Customs classification number
- Approximate duties to pay on the goods
AI-powered software, developed by companies such as Deep Cognition, can ensure that data flows smoothly and automatically from one activity to another. This reduces human error and frees up resources for more value-adding activities by allowing businesses to function in an integrated digital environment as well as submit customs declaration forms faster and more accurately. The technology can process descriptions of goods, cross reference these against tariff nomenclatures, work out the duty payable and update transport management systems, whilst reducing time and resources.
AI is set to significantly increase the efficiency of major organizations. However, fears over the unintended consequences of AI experimentation are likely to become a recurring topic of debate over the coming years.
Within supply chains, more effective allocation of assets in response to demand peaks and troughs will reduce costs. In addition, the prospects for using AI in a creative manner will allow organizations to solve problems in different ways.
As AI applications become more advanced, this will eventually create a self-correcting supply chain which is adaptable and responsive to changing circumstances. Combined with strategic analysis, this could result in an evolving system that is able to recreate itself to support different requirements.