The coming Age of Autonomy: How robotics will reshape the future of logistics

We have long awaited the introduction of robotics into the supply chain. So long, in fact, that some people began to believe that autonomous vehicles and robots would never be anything more than a science fiction-style experiment. But that was because, until recently, robots were stationary, simple-functioning, expensive, blind, and relatively unintelligent. Today, thanks to the convergence of machine learning and artificial intelligence, big data, mobility, and advanced sensors, the autonomous revolution is on the horizon. The robots of the future will be smarter, lighter, and easier to program. They will also be more flexible and affordable. Indeed, it's no longer a matter of "if" robots will be operating delivery vehicles, distribution centers, and factories but "when." The scenario shown in Figure 1, where robots and autonomous vehicles play a role in virtually every step in the supply chain, from field to final consumer, is no longer far-fetched. Each discrete component of this scenario is already being created and tested; it will not be long before they are connected, and a fully autonomous, end-to-end supply chain becomes a reality.

Instead of seeing this technology as a threat, business leaders need to view it as an opportunity.  Autonomous vehicles have great potential for alleviating some of the growing pressures on supply chains to respond faster and more nimbly to customer demand. Because autonomy allows machines to make decisions and act without human intervention, it will help enhance supply chain speed and productivity.

Article Figures

[Figure 1] Fully autonomous, end-to-end supply chainEnlarge this image

[Figure 2] Autonomy landscapeEnlarge this image

[Figure 3] Framework for selecting use casesEnlarge this image

[Figure 4] Selection hierarchy for autonomous vehiclesEnlarge this image

Without a doubt, this revolution will have profound implications for supply chains and for the practice of supply chain management. It's important that business leaders start taking steps now to understand autonomous technology and plan for the day when it provides a viable option for their business. As part of that planning process, companies need to consider two macro questions:

1. How could autonomous technology radically change our company's operating model and value proposition? Could it lead to breakthrough supply chain performance improvements?

2. Can our company anticipate and participate in the new "ecosystems" that will develop around autonomous vehicles? These complementary functions and services will include such things as vehicle manufacturing, financing, and insurance; traffic management technology; workforce training;and parking, fuel, and maintenance, to name just a few.

These two questions are not just complementary; they are inextricably intertwined, and one cannot be answered without considering the other. (The following infographic shows the complete autonomous vehicle landscape and ecosystem.) This article, however, will primarily consider the first opportunity—new value propositions and operating models in a supply chain context. It will also briefly introduce the business ecosystems that will emerge as companies think through, experiment with, and adopt autonomy. This latter opportunity will be discussed in greater detail in a future article.

Promise and possibility

The term "autonomous vehicle" refers to any machine that is able to move, either on its own as directed by preprogrammed software or under remote control by a human operator, to complete one or more tasks. Figure 2  identifies the remarkable—and continually growing—number and variety of autonomous machines that support logistics and supply chain activities. Some of these vehicles are already in widespread use, while others are still in the pilot or development stage. Autonomous vehicles may be huge in size (aircraft, container ships, and highway trucks), relatively small (delivery drones and robotic picking arms), or something in between (forklifts, automated guided vehicles and order-assembly robots).

As Figure 2 shows, autonomous machines in the supply chain operate in four general environments:

• In the air, usually by drone
• On the ground, usually by vehicle
• The "last 100 feet," where cargo or merchandise is transferred to the user
• "GPS-denied," within the four walls of a warehouse, retail backroom, or even a kitchen, where tracking and guidance via global positioning systems and other technologies is not involved

These autonomous systems have the potential to revamp the supply chain and create breakthrough performance improvements. They will do so in at least three general ways:

• By helping controlled environments, such as warehouses, more efficiently manage and process products through autonomous inventory management, including positioning, routing, sorting, storing, tracking, and packaging.
• By increasing transportation efficiency through autonomously moving goods from factories, retail centers, and distribution centers to homes, offices, or other locations of choice.
• By increasing information gathering, sharing, and integration to ensure an optimally performing system.

From an internal operations perspective, every pathway and node of the supply chain can be rethought with an eye toward utilizing autonomous vehicles to achieve radically better efficiencies and cost reductions. Point-to-point long hauls, local routing, yard management, and cycle counting in the distribution center (DC) and store backrooms are all opportunities that offer significant returns on investment (ROIs) in autonomous assets. For example, recently UPS stated that if drones cut just one mile per day from each of its drivers' delivery routes, it would result in US$50 million per year in fuel savings.

From an external, customer-facing perspective, autonomous vehicles could help retailers and delivery companies meet customers' delivery requirements based upon their preferences and economics. For example, "smart" carts, such as Canvas's robotic carts, could help labor-strapped retailers better serve customers who want to pick up their orders in store. A customer could walk in at any time and scan the order information at a kiosk; a cart would immediately bring the order to the customer, eliminating the need for appointments or waiting for an employee to become available to look up the order and retrieve it from a backroom. Other "goods-to-person" technologies include mobile delivery bots with a storage container that consumers must open to retrieve their orders—Starship's delivery bots, which can navigate city streets right to a consumer's designated pick-up location, are in pilot testing now. And, of course, there are UPS' ultra-fast, deliver-anywhere drone package deliveries, which also are in a proof-of-concept pilot now and could soon be in widespread use.

The scope of potential applications for autonomous vehicles in the supply chain is vast. Here are just a few general scenarios where autonomous vehicles offer the greatest promise:

1. Data gathering and analysis—During natural disasters, it can be difficult to obtain up-to-date data on conditions without putting human lives at risk. Drones could continually gather that information and relay it in real time to human decision makers.

2. Capacity challenges—The U.S. transportation market is facing significant capacity challenges due in part to the driver shortage. Self-driving trucks with no restrictions on daily operating hours could alleviate the driver shortage and relieve some of the capacity crunch.

3. High labor costs—Last-mile delivery is often one of the most expensive segments of an e-commerce transaction because it requires lots of labor. By reducing the need for operators, self-driving delivery vans could bring down the cost of last-mile delivery.

4. Speed and efficiency—Autonomous scheduling and dispatching of on-demand deliveries and transportation would reduce lag times and enable faster, more accurate responses than those that could be generated manually.

5. Mobility and safety—Autonomous vehicles could make it feasible for disabled employees to get to work, thereby expanding the labor pool for warehouses and other supply chain jobs. Drones could also carry out dangerous work such as inventory counting in the high racks that are increasingly common in today's warehouses.

As these scenarios suggest, the introduction of autonomous technology will require companies to completely rethink their operating models and value propositions. For example, if autonomous vehicles are introduced to the freight transportation industry, it would mean that human assets would no longer be the limiting factor. The lifting of this constraint means that new, creative operating models would become possible. Without human drivers and hours-of-service limitations, assets would become usable 24 hours a day. As a result, a consortium of shippers might, for instance, collectively bid on a transportation asset, and each would fractionally own the asset or pay a specified share of the cost of the move. If that model arose, companies would also have to remodel their operations to consider how they could structure service-level agreements to keep that container or trailer filled at all times while using and paying for a fraction of it. They would also need to consider whether there is some monetary advantage to be gained from the large amounts of data created by an autonomous vehicle.

Part of an ecosystem

In order for companies to successfully and safely implement autonomous technology into their supply chain, they must also address a number of applications, issues, and requirements. These adjacent concerns include such considerations as how the technology will change the customer interface, how will employees interact with the technology, regulatory compliance, operating rules, data integration, and much more.

One crucial concern that must be addressed is how human beings will interact with the system. Various researchers and organizations have different perspectives on how autonomous systems will or should be designed. Some, like Google's autonomous car division, have adopted an approach that keeps humans completely out of the loop. Others argue for keeping humans in the system and letting autonomous systems and humans jointly make decisions.

The reason for keeping humans involved is so that they can manage novel or unexpected situations. But if companies do decide to involve human beings in the system, they need to carefully and purposefully lay out what their roles should be.  Some research has shown that humans are not effective monitors of autonomous systems because they get bored or complacent if they aren't part of the system's operation.¹Â For safety, it will be necessary to ensure that any humans involved stay engaged.

Additionally, there will be a need to consciously balance employees' concerns about automation with the requirements of the business.Recently the Teamsters trucking union demanded that UPS agree not to use drones or self-driving vehicles to automate deliveries. The union worries that drones could reduce the need for drivers and thus eliminate jobs. However, UPS and other delivery companies are struggling with growing delivery volumes driven by the rise of digital commerce, a situation exacerbated by a worsening shortage of truck drivers. The challenge for these companies will be to use technology to mitigate the labor shortage without alienating employees and potentially triggering a strike or other labor action.

Another human-related consideration will be how a fully autonomous system affects the customer service experience. Generally, delivery of the product to the customer is the crucial point in the supply chain where the company receives the most feedback on the end user's experience. Utilizing automation, though, removes points of "human interaction." People may not convey their feedback and feelings to a robot accurately, or the robot may not be able to interpret the user's true feelings. Until robots can understand body language, facial expressions, and the difference between sincere speech and sarcasm, businesses will have to ensure they have a system in place to stay connected with their customers.

It's also critical to consider how autonomous vehicles integrate with and influence other functions, policies, and technologies. So far, most enterprises look at autonomous technologies in isolation. Yet even if the activity performed by the autonomous vehicle is geographically limited, it will impact other operational and administrative areas. (For an example, see the sidebar on "The complications of implementing warehouse drones.")

Another related consideration is what technical and operating standards will autonomous vehicles need. Autonomous vehicles depend on a composite of technologies including: computer processing, sensors, battery technology, the Internet of Things, machine learning, wireless communication, and software applications and algorithms—just to name a few. How do we get them all to "play well together"? How we answer that question will deeply influence our approach to autonomy in the future. Should there be an open platform that companies can plug into and create their own apps for specific purposes? This would be analogous to the way many apps have been developed based on Apple's iOS operating system.

In my view, a common operating system for autonomous vehicles that would facilitate "plug and play" applications will be critical. Such an overlaying architecture would prevent vehicles—whether drones, trucks, delivery bots, or anything else—from operating in conflict with each other. Common standards would provide the ability to gather information from any kind of sensor, make sense of that data, and respond in a way that is predictable and conforms to established norms. For safety's sake it will also be important to build in algorithms that govern how a vehicle will react or behave in specific situations or environments.

At the same time, companies will also need to keep on top of the new complementary market opportunities that are formingaround autonomous vehicle technologies. This "ecosystem" of business opportunities will be discussed in more detail in an upcoming article, but briefly, they include:

• Manufacturing—designing and manufacturing autonomous vehicles at scale
• Traffic management—developing the technology and monitoring the safe operation of vehicles
• Parking and toll infrastructure—providing parking locations for autonomous vehicles and developing toll-collection systems
• Inventory, maintenance, and fueling—managing electronics and parts inventory for autonomous vehicles; training and providing skilled labor for repair and upkeep of autonomous vehicles; and developing and operating fueling stations for the vehicles
• Asset financing—providing appropriate financing for autonomous vehicles
• Insurance—creating a new class of vehicle insurance that is not based on human behavior
• Data storage, processing, and monetization—providing storage and processing capacity for the enormous amounts of data that will be created by autonomous vehicles as well as managing governance and access to that data

Gaining acceptance

There will be barriers to adoption of autonomous systems, most prominently the difficulty of gaining acceptance. For example, autonomous systems are likely to operate in conjunction with legacy systems, and if they are to be accepted, companies must ensure that autonomous vehicles can seamlessly operate in a "mixed" environment. Moreover, the cybersecurity of autonomous systems will have to be assured.

Safety is another concern that will affect the acceptance of autonomous vehicles. They will have to be proven safe before they can be widely adopted. This does not just apply to new activities; it's important that autonomous cars and drones do not negatively impact the safety of existing operations, too.

In traditional transportation and distribution systems, humans have demonstrated the value of their contributions under off-nominal (abnormal) conditions or when contingencies (chance events) occur. The autonomous systems will have to show that they can effectively manage any off-nominal conditions, contingency operations, and novel situations. It is therefore critical to ensure that the autonomous systems are really ready for prime time and can safely and accurately operate in complex environments.

Given all those concerns, it's not surprising that most company leaders have not considered whether and how to adopt autonomous vehicles. Nearly two-thirds (64 percent) of the executives who participated in a Techpro study, for example, said they have not given any thought to applying autonomous technologies in their companies. Only 21 percent said they are beginning to craft a strategy, 13 percent are deciding where to apply autonomous technologies, and less than 1 percent are already active in this space.²

However, those holding back should not wait too long for a number of reasons. First, a company that does not investigate the potential of autonomous vehicles puts itself at a competitive disadvantage. Nobody wants a direct competitor to gain an insurmountable lead, especially in new technology where the first mover often has a distinct market advantage. But there are other reasons that are specific to autonomous vehicles. For one thing, companies that get into the autonomy game early on will have an opportunity to help shape policies and legal frameworks under development by regulatory agencies like the U.S. Federal Aviation Administration (FAA). Influential companies such as Amazon and Google have differing views on whether the architecture of airspace for autonomous vehicles should be closed (and therefore tolls could be charged) or open to every operator. If others don't get involved and make themselves heard, just a few companies will influence government policy. An intelligent stand can only be taken through a fact-based experimental process. For another, specific operating concepts, even in the testing phase, must be certified as safe and be approved by the regulating agency before a pilot can proceed. Whenever a variable changes, companies have to go through the rigorous certification process again. Certification takes a long time—18 to 36 months for drone applications regulated by the FAA, for instance. Innovators should therefore move ahead as quickly as they can to get approval for testing.

Get ready for the future

There's no doubt that it will be a daunting task to adopt an autonomous system to replace a current practice or to perform new tasks. In order to ensure that an autonomous system will work effectively, companies should take the following steps:

1. Clearly define the goals and tasks for which an autonomous system is being considered. What will it take for autonomous vehicles to become an integral part of our mobility landscape? They must be applied in theright business context to solve the right problems. In other words, a critical success factor will be a clear value proposition for the company and its customers, and in some cases for society at large. Without a clear and viable value proposition, autonomy projects will lack support and are at risk of failure.
2. Identify/define what constitutes the successful completion of that task. Any application of autonomy must also provide some economic benefit to the provider of products and services. Consider that customer-facing autonomous solutions must integrate, both operationally and in terms of information technology, with supply chain systems further upstream. To accommodate autonomous deliveries, companies will have to make process and infrastructure changes. The question is whether that can be done in a way that reduces costs and/or improves profitability.
3. Identify all possible off-nominal conditions, contingencies, and challenging edge cases (problems that occur at extreme operating parameters) that the autonomous system must address. Bad weather, for instance, could push a drone off-course, causing it to land outside of its targeted landing zone. Malfunctions—whether mechanical or caused by software glitches—or unexpected conditions such as temporary obstructions could cause a collision or crash, causing damage to the payload and, potentially, to humans. These are just two of many possible challenges vehicle users will have to identify and plan for.
4. Test the autonomous system in a simulation environment with many nominal and off-nominal conditions. To ensure that autonomous vehicles and their supporting technologies perform safely and correctly, it will be critical to test them in both normal, expected conditions and abnormal, unexpected situations. Neither will be easy, but the latter involves potential risk to life and property; it would be unacceptable and unethical to take such risks. The safest way to test is to use a simulation program (similar to a flight simulator) to re-create real-world scenarios and use that information to make improvements. In other words, virtual testing of autonomous vehicles should augment and "stress-test" the physical testing.
5. Prioritize projects. Create a methodology for prioritizing which sort of autonomous vehicle projects to focus on first. One way to structure it might be a "crawl-walk-run" approach, starting by adopting autonomy for simple tasks where the risk is lowest and there are fewer variables to control (for example, a single vehicle traveling a repetitive route in a sparsely populated area), and then moving to more complex tasks with more variables (multiple vehicles traveling variable routes in an urban, congested area).

Figure 3 provides a framework of seven considerations that can be used to assess the complexity of use cases for air and ground autonomous vehicles, progressing from low to high complexity. The seven considerations or categories of complexity are as follows. Operation type:Autonomous vehicles that only interface with the operator are much simpler to manage and  have fewer risks and liability concerns than those that interact with customers. Topology type: It is easier to manage and implement an autonomous vehicle operating in a rural area than one in a congested urban setting. Control type: When autonomous vehicles are operated within visual line of sight (VLOS) they are easier and safer to control than when they operate beyond line of sight (BVLOS). Payload:Drones or other autonomous vehicles that operate without a payload are easier to operate than those that are carrying a payload. Carrying a payload reduces the drone's range and exposes the user to such liability issues as material loss and injury to humans. Functional use: The more functions and technologies that are integrated into the vehicle, the more difficult it is to manage. Movement type: It is much easier to manage an autonomous vehicle that travels along a point-to-point route than one that traverses multiple congested, heavily populated urban areas. Asset in motion: It is simpler to manage and instrument a traffic-control system for one vehicle than a multi-asset model requiring integration of ground and air systems.

Another way to prioritize autonomy projects is to think about what problem a project would solve and rank it in order of its importance to the company and/or its customers. For example, a company's most important reasons for pursuing an autonomous vehicle project might be to improve safety and access to critical information. Improvements in speed, productivity, and cost, while important, could be lower priorities. Figure 4 shows six potential areas that companies may wish to improve and provides examples of potential autonomous vehicle implementations that would help with each.

6. Ensure that you have a plan for managing data collection, storage, and analysis. Every autonomous vehicle constantly produces enormous amounts of data. This creates three issues. The first is how to collect all that information and use it, without latency, to make real-time decisions and take real-time or near real-time actions. The second is how to determine which data should be stored for later use. Out of that huge stream of data, which signals are important to keep, and which can be discarded? The third is how to distribute the storage of so much data. There will be too much data to keep in one location—and from a risk management perspective, it would be unwise to do so anyway. But when forensic research becomes necessary—perhaps to recreate an accident and identify its causes—it will be critical to know where all the data is sitting and how to access it.

We are just beginning to scratch the surface of how autonomous vehicles will impact logistics and supply chain operations. For example, in the future, autonomous vehicles will play an important role in data gathering, processing, and sharing. Processing advances in sensors, machine learning, and artificial intelligence are already pushing computing farther out in a supply chain. Autonomous vehicles could not only function as a mobility system, but they also could serve as both a local storage device for the relevant sensor data and as a "mesh network" for processing it—an example of "edge computing." This newly data-rich information infrastructure will help enterprises to enable fluid alignment (and realignment) of resources and workflows through a major shift in their network and supply chain architectures.

The successful businesses of the future will be those that are able to adapt to rapid changes in production, sourcing, and distribution, as traditional strategic boundaries will be upended. It will take foresight and a willingness to focus on the future to place the strategic bets on flexible and agile technologies that will position a business for success five to 10 years from now.

Because supply chain organizations encompass warehousing, transportation, and associated technologies, and because they touch nearly every corporate function in some way, they are well positioned to lead the adoption of autonomous technologies. They must be prepared to act as the connecting thread among internal interests such as information technology and human resources and be able to help colleagues understand opportunities and use cases. The ability to play a major role in change management and developing a cross-functional governance model will be essential, as will a commitment to supply chain innovation. It will not be easy—adopting a game-changing, radically new technology never is. To achieve such challenging goals will require supply chain innovators to move aggressively yet pragmatically. It's best, though, to invest in an autonomy project only after you've determined that it's a strategic fit for your company and that it will be a core activity of the organization.

Notes:

1. R. Parasuraman and D.H. Manzey, "Complacency and Bias in Human Use of Automation: An Attentional Integration," Human Factors:The Journal of the Human Factors and Ergonomics Society, October 20, 2010.

2. Tech Pro Research, "Autonomous transportation research: Predicting impact on industries, companies, and personal lives," February 2018, https://www.techrepublic.com/resource-library/whitepapers/special-report-tech-and-the-future-of-transportation-free-pdf/