Innovations in Last Mile Logistics

AI and human intelligence are a potent blend in route planning and optimization.


innovations in last mile logistics

Artificial intelligence (AI) is currently in the hype stage of new technology adoption; some proponents see it as the answer to many problems, including, for the supply chain industry, last mile logistics. While AI does have potential for resolving some last mile issues, it is just one part of an overall solution. It will take a combination of technology and human intelligence to take last mile delivery to the next level.

Logistics companies are generating massive amounts of data from their operations each day, and it can take them weeks or months to analyze data and resolve problems. With today’s powerful computers and AI, however, they can process this information in real time, gaining actionable insights into their operations. With AI, they can be more proactive in solving problems and/or can identify opportunities for improvements in their operations that will provide them with a competitive advantage.

Descartes uses AI to help customers configure its route planning systems for their specific needs. After customers answer a series of questions about their business objectives, the AI, drawing upon the logistics expertise which Descartes has embedded into the system, can recommend settings that will optimize routes to meet those objectives.

The customer can then run trial scenarios with those settings to test the results. “We created a closed loop environment where you get better and you can also understand more of what the implications [of a scenario]are before you actually ever launch it,” said Chris Jones, Descartes’ executive vice president of industry and services.

They also take advantage of machine learning (ML) technology to improve route efficiency. The system’s AI evaluates GPS delivery data to determine how long a stop to deliver an order of a certain size will take. If the results show that drivers take just four and a half minutes rather than the five minutes allocated by the planning system, the company can reduce stop durations and possibly add an extra stop to that route, increasing productivity.

“All of that will happen in the background. Machine learning will do the performance analysis that people would do on their own but have never done very frequently because it’s a lot of work,” said Jones.

Transportation management systems that incorporate AI are moving away from simply providing information to providing prescriptive analytics, said Bart De Muynck, strategic advisor at Mojix. “They can say, based on what has happened, that you should think of doing one of these three options.” AI can suggest a dynamic response to a problem; if a product is stuck on a ship waiting to be unloaded, the AI system could sift through inventory data and recommend that the delivery to the customer be completed on time by taking the product from a nearby store.

Data Dependent

The quality of the insights that AI technology provides depend on the accuracy and timeliness of the data that it processes. That can present problems for companies that haven’t kept their data accurate and current.

AI relies on the data that GPS technology generates at every step of last mile loading and delivery activities. It also needs access to master data, which includes information about customers and products (weight, dimensions, how it needs to be stacked or stored, etc.). Master data is often outdated and usually needs cleansing and updating, according to De Muynck.

Machine learning can be helpful in this cleanup process. A company receiving a delivery or arranging a pickup may list the address at their main entrance, but the actual stop may be at a side entrance around the corner. After analyzing the GPS data, the AI could recommend a change in the delivery address to reflect where the drivers actually stop. That change has the benefit of making the delivery route plan more accurate, and assists drivers who are unfamiliar with the route.

There are limits on what AI can do with even the most current data, however. For example, it’s not currently feasible for AI to use accident reports and other real-time traffic data to change delivery routes on the fly. One problem is the limited number of routing options. If an accident closes an interstate, AI systems might suggest an alternate route that quickly becomes just as jammed up as the interstate, or one that has commercial traffic restrictions.

Jones believes that predictive road speeds that use AI technology are being incorporated into routing systems and will become more common in the future. Combining AI with historical data about travel speeds on roads at various times of day, such as rush hour, could build more realistic times into delivery routes.

Guided by Human Intelligence

De Muynck said that 80% of what people do today in last mile logistics could be automated by using AI. But that doesn’t mean AI will replace dispatchers; it will help them work more efficiently and be more productive.

“It’s really about how we make everybody as good as the best planners out there,” Jones said. “We know from experience that even within the same company, some people consistently produce much better results. So, if you can take [their]practices, their learning, and automate it, then what you’re doing is taking those best practices and giving them to everybody so we can raise them all up. For us, all the forms of AI that we’re involved with are really more enablers for improving the overall quality of planning.”

AI will change what humans do. People now have to spend much of their time generating reports and doing basic analysis, but with machine learning performing these basic tasks, humans will be free to do deeper or more complex analyses. “We think it’s going to help people be more productive,” Jones said.

Humans also determine what data gets fed into an AI system in order to meet business objectives. The technology can’t decide on its own to expand the number or types of carriers in its database so that a company can increase capacity, cut costs or improve delivery times. The AI also can’t understand the long-term implications of getting a delivery to a particular customer at a certain time.

Click here to read the full feature.