부산시청 도서요약
국내도서 요약
국내도서 프리뷰 해외도서 프리뷰
글로벌 트렌드 미디어 브리핑스
   글로벌 트렌드 

책표지






  • Driverless Trucks Do Not Need Every Road
    - The economics of autonomous freight starting on highways and inside ports

    An empty truck cab prompts an immediate question: are drivers about to disappear? Yet the first thing a transport company needs to examine is less the empty seat than the two places where that truck starts and finishes its journey. The business case for autonomous freight first emerges from the ability to move goods safely along a defined route and hand them over precisely to the next worker, rather than the ability to drive on every road.

    [Key Message]
    * Autonomous freight first becomes a business on routes where operating conditions are easier to define, such as highway journeys between hubs and repeated movements inside ports. Selecting a limited route is the starting point for safety validation and operating cost calculations.

    * An announcement that autonomous transport has begun must be distinguished from operation with nobody inside the vehicle. A safety driver on board, a non-driving observer, driverless operation, vehicle delivery, and on-site validation represent different stages.

    * Even when driver costs fall, new expenses arise for hub operations, remote assistance, sensor care, maintenance, and emergency response. Companies need to compare the cost and on-time performance of the entire freight journey rather than the price of a single vehicle.

    * A driver's duties extend beyond driving to load checks, vehicle inspections, paperwork, and customer communication. Automating driving requires deciding who will take over these tasks, while the content and compensation of frontline work also change.

    * Korea first needs to select routes and tasks for validation rather than draw a blueprint for driverless trucks nationwide. Pilots become repeatable services when actual freight demand, permitted operating conditions, work handovers, and emergency response are designed together.

    ***

    An Autonomous Truck's First Customer Is a Defined Route
    Imagine a truck leaving a logistics center, traveling along a highway, passing through urban side streets, and entering a customer's warehouse. For a person, this is a single transport assignment. From an automation perspective, it is a bundle of fundamentally different problems. Merging onto the highway, passing through roadworks, negotiating a narrow entrance, reversing into a parking space, and interpreting a loading dock worker's gestures all have to be resolved before one delivery is complete.

    Trying to solve these problems simultaneously expands the scope of development and validation. Dividing the journey creates alternatives. A local driver can move the freight to a hub near the highway, an autonomous truck can carry it between hubs, and a driver at the destination can handle the final stretch. Different driving arrangements are connected within a single origin-to-destination transport service.

    The hub-to-hub model proposed by Volvo Autonomous Solutions follows a similar direction. Autonomous trucks are intended to handle the highway portion, with people responsible for local transport. The key is to go beyond supplying vehicles and design hardware, software, and operating services together. What a carrier wants to buy is less a special truck than the capacity to deliver at the promised time.

    An important concept here is the operational design domain (ODD): the boundaries specifying the roads, speeds, weather conditions, and times of day for which a system is designed to operate. Automated driving safety guidance from the US National Highway Traffic Safety Administration also treats the definition of this domain, and responses that reduce risk when the system moves beyond it, as major considerations. Drawing a route on a map does not by itself establish its operating conditions.

    The same road presents a different test on a clear day and on a night of heavy rain. Faded lane markings, temporary restrictions, and a disabled vehicle on the shoulder change driving decisions. Any claim that a truck can use a particular highway must therefore explain under what conditions it can do so and when departure is postponed. Rules for suspending operation are necessary before the service can enter a real dispatch schedule.

    Choosing highways first does not make the challenge easy. At high speeds and with heavy vehicles, even a small error in judgment can have serious consequences. A repeated route helping validation and that route being safe are separate propositions. Commercializing limited routes does not mean risk disappears; it means the conditions requiring validation and the responsibilities of operation can be defined concretely.

    Seen this way, the initial autonomous driving product is closer to a service connecting two hubs under specified conditions than a driving capability that works everywhere. Just as airlines define routes and operating criteria rather than tackle every airport and every weather condition at once, carriers build operating capabilities lane by lane. Narrowing the scope of application can make the service easier to explain.

    Texas's Empty Driver's Seat: Why the Whole Announcement Matters
    In May 2025, Aurora announced the launch of commercial driverless freight transport between Dallas and Houston, Texas. It said regular customer freight deliveries had begun and that, at the time of the announcement, its trucks had covered more than 1,200 miles without a driver. This was a significant moment in the transition from technical demonstrations to actual customer transport.

    But expanding that scene into a claim that American trucks now run without people would depart from the facts. One company launching a service on a particular lane does not mean routes nationwide and all weather conditions have been validated. Starting customer deliveries does not itself demonstrate the cost competitiveness or profitability of that operation. Commercialization can be a starting line without constituting a complete report card for the business model.

    In mid-May that year, Aurora explained that a person would again occupy the driver's seat. Its partner PACCAR had made the request in connection with certain prototype components in the vehicle platform, and the company agreed to move an observer who had ridden in the rear seat on some trips to the front seat. Aurora said the observer would not drive and that the system would remain responsible for driving and any necessary safe stop. The case also shows why a person sitting in the driver's seat cannot automatically be assumed to be driving.

    The launch of second-generation driverless trucks followed in July 2026. The company said it then had ten driverless routes across the US Sun Belt and introduced the new vehicles as operating without a person behind the wheel. This is operating information disclosed by the company, not a measure of market-wide adoption or an independent safety evaluation. The operating conditions of an initial launch and subsequent vehicles can change, and that distinction needs to be read alongside the announcement.

    These cases also clarify other announcements with similar headlines. The Texas autonomous transport operation announced by DHL and Volvo in December 2024 initially included a safety driver. It had entered commercial logistics by carrying actual freight, but its operating conditions differed from those of transport with nobody inside the vehicle. There is an important gap between using autonomous driving technology and removing the driver.

    Investment reviews and purchasing meetings should turn that gap into questions. Is anybody inside the vehicle? If so, do they intervene in driving? Is this a regular service paid for by a customer, or a test operation? Do the announced vehicle numbers refer to trucks delivered, trucks that have completed validation, or trucks actually assigned to freight runs? A few short questions reveal operating maturity more effectively than polished promotional language.

    A photograph alone rarely supplies the answers. An empty cab attracts attention, but a carrier's contracting team needs operating hours, service exclusions, incident procedures, and commitments for substitute transport. Assessing progress in autonomy requires reading the scope of the service's responsibilities rather than simply the technology's name. This standard recognizes positive achievements while distinguishing them from expectations that remain unvalidated.

    A Port Is a Small World, but Not a Simple One
    Where highways address long movements between hubs, ports address repeated movements within a defined space. Moving containers between pieces of equipment and setting them down in designated positions allows relatively clear definitions of origins, destinations, and movement instructions. Operators can also exercise some control over outside vehicle access and workers' movements, making it easier to design the scope of automation.

    Fences do not make obstacles disappear. Large machines move, containers block sightlines, and human-driven vehicles may share space with automated equipment. Precise reversing and positioning are also required. The advantage of a port is not that every situation is easy, but that responsibility for adjusting site rules and traffic priorities is relatively clear.

    APM Terminals Maasvlakte II at the Port of Rotterdam illustrates this distinction. In April 2026, the terminal and its partners announced the arrival of five additional electric automated terminal tractors, bringing the number on site to ten. The planned fleet is thirty, but the additional vehicles had completed factory testing and were awaiting site acceptance testing. Vehicles that have arrived on site should not immediately be counted as vehicles in full operation.

    The same announcement explained that the first five vehicles had completed site acceptance testing under controlled operating conditions in November 2025. Validation covered mixed traffic, narrow-lane driving, lane changes, and reversing into transfer points. Progress in port automation is revealed not only by vehicle counts but also by which maneuvers have been validated in which environments.

    Truck travel time alone does not determine a port's productivity. If a crane is not ready or the next job instruction arrives late, a vehicle waits even after reaching its destination. An automated tractor arriving one minute earlier does not accelerate the whole operation by the same amount if the equipment receiving its container will not move for another ten minutes. Vehicle performance and terminal throughput need to be examined together.

    That requires aligning the format of job instructions. Which container goes where, whether the destination is available, and who sends the completion signal all need to connect. Small confirmations once handled by radio become data and approval procedures in an automated environment. Gaps in instructions and handovers may become apparent before the driving system is even introduced.

    In its 2025 discussion of autonomous freight, the Netherlands Organisation for Applied Scientific Research (TNO) also emphasized starting in controllable environments. It highlighted the need to standardize and digitize other driver duties, including connecting trailers, handling transport documents, and checking loads. From this perspective, ports test how clearly surrounding work has been designed as well as how well autonomous trucks drive.

    Remove Driver Costs, and Other Expenses Come into View
    The quickest way to explain autonomous trucking economics is to subtract driver costs. A quick calculation is not necessarily an accurate one. Companies must distinguish between work that disappears, work transferred to other people, and work converted into equipment and service expenses. Particularly during early deployment, a small number of trips must absorb the fixed costs of vehicles and the operating system.

    The highway hub-to-hub model adds the operating costs of departure and arrival hubs. Freight or trailers must be handed over, vehicles checked, and parking and queues managed. A conventional human-driven vehicle may travel directly to its destination, while the automated model may require an additional hub stop. Benefits on the long trunk segment must offset extra handover costs and time for the full journey to make economic sense.

    Support outside the vehicle also costs money. Operations may require people to clean and inspect sensors, a maintenance network to recover disabled vehicles, emergency responders, and a control organization to support operation. Communications, software, insurance, and service contracts also need review. Removing one line of driver payroll introduces several other cost categories, so the change in the total must be examined.

    Numbers make the issue clearer. The following is a simple hypothetical calculation, not an actual company's performance. Suppose the relevant daily cost of conventional transport is 100 and automation removes 30 of that cost. If new vehicles, software, and support operations add 18, while hub handovers and additional waiting add 8, the net saving is only 4. If the new costs total 35, the overall cost instead increases by 5 under otherwise identical assumptions.

    This calculation shows that what is included in costs matters more than the isolated saving of 30. Real comparisons need consistent treatment of vehicle depreciation, fuel or electricity, empty running, canceled trips, and incident response. Excluding from the new model an item included in the existing one can make the comparison look favorable. Costs left outside the contract are still borne by somebody within the overall transport operation.

    Utilization matters too. Fixed cost per movement changes when the same hub is used twice a day rather than twenty times. Yet a vehicle's ability to operate longer does not automatically increase revenue. Without return freight, or when the customer's loading dock is closed, empty movements or waiting result. Machines not needing sleep and customers accepting goods all night are different things.

    An initial deployment lane should therefore not be selected on distance alone. Companies need to consider freight in both directions, repeatable departure times, hub access, handover duration, and the likelihood of moving outside permitted operating conditions. Comparing cost per completed loaded movement, the share completed within the promised time, and the number of additional support requests makes the judgment more concrete. Two lanes of similar length can perform very differently because of their operating conditions.

    What a carrier sells to its customers also becomes important. It can sell vehicle hours, movements between hubs, or a contracted amount of transport capacity. Each arrangement allocates the risks of waiting, cancellation, and insufficient demand differently. The contractual distribution of costs can influence adoption as much as technical performance.

    Does Somebody in a Control Room Drive the Truck Instead?
    When a control center enters the autonomous trucking story, people may picture a giant gaming room: staff sitting before multiple screens, each steering a distant truck. Remote assistance and remote driving, however, need to be distinguished. Providing a vehicle with situational information or helping it choose a route differs in both role and requirements from directly steering and braking it remotely.

    Suppose roadworks block the usual entrance. In one arrangement, support personnel might identify an alternative entrance and communicate it, while the vehicle drives using its own system. In another, a person might drive a specific stretch remotely. What is possible depends on the design and permitted operating conditions of the vehicle and service. Merely saying that a control center exists does not explain the system's limits or the nature of human intervention.

    For driverless operation, the vehicle's response to a lost communications connection also matters. The existence of remote assistance does not mean every moment of safety judgment can be entrusted to that connection. Guidance from the US National Highway Traffic Safety Administration sets out the expectation that, at higher levels of automation where a driver cannot intervene, the system itself should be able to transition to a condition that reduces risk. This is voluntary safety guidance, not a safety certification for an individual vehicle.

    Claims that a vehicle can stop also need detail. Where does it stop, how does that affect following traffic, and what information reaches the road operator and response team? A stop inside a tunnel is not the same situation as a stop in a large hub parking area. Emergency response should be a procedure connecting roads and operating organizations, rather than merely the final line of a driving algorithm.

    Control staff productivity cannot be judged solely by the number of vehicles visible during normal operation. Sudden heavy rain in one area or a major road closure can cause multiple vehicles to request assistance simultaneously. Even if one person normally handles many vehicles, responses may slow when requests surge. Support staffing and dispatch limits therefore need to be designed together.

    It also matters who has authority to postpone departure. If a control employee identifies a risk and suspends a trip but is then held responsible for the delay in a performance review, judgment can be distorted. Operating rules should prevent delivery speed targets from conflicting with the authority to keep operations within safe boundaries. This is an organizational design issue to consider during deployment, not a claim about a verified problem at any particular company.

    Customer contracts need the same principle. When a trip cannot depart because conditions fall outside the operating domain, responsibility for arranging substitute transport and paying additional costs must be assigned. Timely suspension is necessary for safety, but the customer still needs the freight. Autonomy becomes an everyday transport product when safety decisions and service responsibilities can be explained together.

    The Steering Wheel May Disappear, but the Driver's Other Work Remains
    Describing a truck driver's day solely in terms of driving time misses the realities of the job. Drivers examine vehicle condition, check loads and securing arrangements, find unloading locations, and contact the relevant staff. They also deliver documents and coordinate waiting order. These tasks vary by company and transport model, but an empty cab does not complete them automatically. This is also why TNO emphasizes standardizing work beyond driving.

    Hub-to-hub automation creates opportunities to divide duties. Different people can handle trunk driving, local transport, vehicle checks, and handovers. But a new division of labor does not benefit every worker equally. Some may gain more predictable times to return home as long-haul driving declines, while others may face changes to their earnings structure or more waiting and short-distance work.

    It is therefore insufficient to end the explanation by saying people will do more valuable work. Companies need to specify which tasks grow or shrink, what qualifications and training are required, and how pay and working hours change. Even if new jobs are created, whether existing workers can take them is a separate question. A workable transition path requires assigning the cost and time needed for retraining.

    Frontline expertise is useful during design too. Drivers and site staff who have worked a lane for years know which entrance causes congestion, when loading queues grow, and where water collects during rain. This is operating knowledge that map data alone may not capture. Incorporating it into the design of handover points and lists of exceptional situations makes problems more concrete.

    Shippers also need to change how they work. Transport is not complete if an automated vehicle arrives but nobody is available to receive it at night, or if work instructions are only conveyed verbally. Arrival notifications, entry approvals, freight acceptance, and error reporting require defined procedures. This is where a carrier's technology deployment meets a customer's frontline work.

    Managers' work changes as well. Problems once solved by calling a driver may be divided among system records, role-based approvals, and response procedures. That does not mean every exception needs complicated paperwork. It is more practical to begin with frequent problems, assign responsibilities and response times, and apply the results to subsequent trips. Standardization for automation should be checked to ensure it does not merely add unnecessary reporting work on site.

    Including workforce transition in the operating plan also improves cost estimates. Assigning existing staff to control, maintenance, and hub work entails training and reassignment expenses; outsourcing entails contract costs. Leaving these as hidden expenses can destabilize staffing plans after deployment. Defining people's roles clearly gives workers necessary explanations and provides managers with proper cost control.

    Korea's First Test Is a Dispatch Sheet, Not a Nationwide Map
    When applying overseas cases to Korea, companies should compare operating conditions rather than simply copy route names. Whether a US hub-to-hub structure produces the same economics under domestic freight volumes, distances, roads, and loading conditions requires calculation. Validation results from internal port operations cannot simply be applied to public roads either. The technology's capabilities must be read together with the extent of control over the site.

    The starting point is a real dispatch sheet. Look for movements that depart at similar times each day, carry consistent types of freight, and allow work procedures at both hubs to be aligned. Then separate highway travel, hub access, and internal terminal movements to decide which tasks to validate first. The approach selects units in which economics and safety can be assessed rather than targeting the entire journey for automation from the outset.

    The first validation task is to record existing work properly. Alongside travel time, companies need to measure pre-departure inspections, handovers, waiting, empty running, and exception handling. Without current costs and on-time performance, improvements from the new approach cannot be established. A comparison requires understanding how well existing operations perform normally and when they break down.

    The second is to document permitted conditions. Specify roads, operating hours, weather and speed limits, and traffic rules inside hubs. Also decide who suspends operations outside those conditions and how the freight continues its journey. This is not a substitute for Korean authorization requirements, but a set of operating design matters to check with the relevant authorities and service providers before implementation.

    The third is to treat handovers as a separate task. Establish who checks the condition of freight, trailers, and vehicles, and at what point responsibility passes to the next person. Automating repeated work does not prevent disputes if handovers remain unclear. Customers also need to agree whether successful delivery means vehicle arrival or confirmed acceptance of the freight.

    The fourth is to establish expansion criteria in advance. Assess canceled trips, additional support, maintenance, and total transport completion time alongside trial mileage and incident counts. A short record without an accident cannot establish safety conclusively. The assessment should also retain which conditions and exceptions have been validated and which situations remain unencountered.

    The fifth is to check whether surrounding operations can support a growing fleet. A single demonstration vehicle can receive intensive support, but entrances, parking, control, and maintenance can become bottlenecks as vehicle numbers grow. Before adding a second vehicle, calculate the support time consumed by the first. An organization that makes a pilot succeed may need a different allocation of resources to operate repeated transport profitably.

    The results of transport automation are clearer in an ordinary working day than in a technology announcement. Scheduled freight departs, trips outside permitted conditions are postponed, necessary assistance arrives on time, and the next worker understands the handover. Customers can commit regular volumes when such days are repeatable. This is the transition from an autonomous truck having moved to a transport service actually operating.

    Driverless trucks do not need every road. Their first routes, stopping conditions, and tasks handed to people need to be defined clearly. The ability to complete one freight movement safely and economically between two highway hubs or two port work points is where autonomous freight's first business begins: in that specific, repeatable round trip.

    Reference
    Volvo Group. Volvo Autonomous Solutions expands its footprint and starts operations in Texas. June 2023.
    National Highway Traffic Safety Administration. Automated Driving Systems 2.0: A Vision for Safety. September 2017.
    Aurora Innovation. Aurora Begins Commercial Driverless Trucking in Texas, Ushering in a New Era of Freight. May 2025.
    Aurora. Updating Our Driverless Operations. May 2025.
    Aurora Innovation. Aurora Launches Second-Generation Driverless Trucks in U.S. to Meet Customer Demand. July 2026.
    DHL Supply Chain. Volvo Begins Autonomous Operations for DHL Supply Chain in Texas. December 2024.
    APM Terminals. APM Terminals Maasvlakte II, Embotech and Terberg expand Automated Terminal Tractor fleet in Rotterdam. April 2026.
    TNO. Self-driving trucks and logistics: Five developments. March 2025.