Automation in trucking, transit, mining, agriculture and sidewalk delivery

Robotaxis attract the most attention, but some of the earliest and most economically compelling uses of vehicle autonomy are elsewhere: on highways carrying freight, on rails and dedicated lanes carrying passengers, in open-pit mines hauling ore, on farms working fields and on sidewalks delivering parcels. Each domain has its own operational design domain, economics, safety challenges and regulation. Several have advantages over urban robotaxis: simpler or more controlled environments, high labour costs or labour shortages, and long operating hours that amortize expensive equipment. This lesson surveys these applications, with attention to Canadian examples, and draws out what they teach about choosing where autonomy works first.

Highway trucking

Long-haul trucking is a strong candidate for automation. Highways are structured, with no pedestrians, controlled access, gentle curves and consistent markings. Driver shortages are persistent in both Canada and the United States, and hours-of-service rules limit how long a human driver can drive each day, while an automated truck could operate nearly continuously. Many developers have adopted a hub-to-hub model: human drivers handle local pickup and delivery in complex urban areas, while automated trucks drive the highway segments between transfer hubs near interchanges.

The challenges are substantial. A loaded tractor-trailer can weigh up to about 36,000 kg in the US and over 60,000 kg in some Canadian configurations, so stopping distances are long and sensors must see far, often beyond 300 m for highway speeds, and must classify small debris in the lane. Trailers vary, with different lengths, loads and dynamics, and the truck must detect trailer sway and handle jackknife risk on slippery roads. Breakdowns, tire failures and roadside inspections need procedures without a driver on board.

At the time of writing (October 2026), automated trucking has begun commercial driverless operation in Texas and the US Southwest. Aurora launched driverless commercial runs between Dallas and Houston in 2025 and has since extended its network, including a roughly 1,000-mile route between Fort Worth and Phoenix, operating some trucks with no one on board. Other companies run driverless trucks in industrial settings such as oil-field logistics. In Canada, Ontario's automated commercial motor vehicle pilot allows trucks to be tested with a qualified driver present, and Canadian companies and carriers have conducted highway trials, but cold-weather operation and cross-border rules remain hurdles.

Winter is a particular test for automated trucking in Canada. Lane markings disappear under snow, plowed snowbanks narrow the road, blowing snow blinds lidar and cameras, and trailers can slide on black ice. Developers working toward Canadian operations combine radar, which sees through snow better than lidar, with map-relative localization to roadside features, and estimate road friction continuously from wheel slip and weather data so that speed and following distance adapt. Until such capabilities are validated, the operational design domain of most automated trucks excludes heavy snow, which in practice confines driverless operation to southern routes or to the warmer months.

Transit

Automated transit is older than most people realize. Vancouver's SkyTrain has run driverless trains since 1986, and Montreal's Réseau express métropolitain, REM, opened its first driverless light-metro segment in 2023. These systems operate at the highest grade of automation for rail, with no staff on board in normal operation, but in a fully protected guideway with platform screen doors or intrusion detection. Their safety record shows what automation can achieve when the operating environment is controlled.

On roads, the picture is different. Low-speed automated shuttles have been tested in many cities in both countries, typically carrying a dozen or so passengers at 20 to 25 km/h (12 to 15 mph) on fixed routes, often with an attendant on board. Trials have shown the technology works in limited settings but also exposed limits: shuttles that stop for every leaf or snowflake detected by their lidar, slow speeds that frustrate other road users, and costs that remain high when an attendant is still needed. Full-size automated buses have been tested in depots, where precise docking and parking are valuable, and on dedicated busways. Automated transit has the potential to serve low-density areas with on-demand service, but integrating it with existing transit and funding it fairly are policy questions as much as technical ones.

Mining: autonomous haulage

Open-pit mining was one of the first industries to deploy driverless vehicles commercially at scale. Autonomous haulage systems, AHS, from major equipment makers operate fleets of ultra-class haul trucks carrying more than 300 tonnes each, around the clock, in mines in Australia, the Americas and elsewhere. Canada's oil sands in northern Alberta are a prominent example: operators such as Suncor and Imperial have converted large fleets of haul trucks to autonomous operation. The environment suits automation: private sites with controlled access, known routes, trained personnel, and an operator who controls the whole system. Benefits reported by operators include improved safety by removing drivers from fatigue-prone night shifts and hazardous areas, more consistent operation, lower tire and fuel costs and higher productivity.

Mine autonomy is not trivial, though. Haul roads change daily as the pit advances, dust and snow obscure sensors, temperatures in northern Alberta fall below minus 40 degrees Celsius, and light vehicles driven by people share the site. AHS deployments rely on high-accuracy GNSS, site-wide wireless networks, central fleet management that assigns routes and enforces separation, and strict site rules: every person and light vehicle carries a transponder, and autonomous zones are controlled. Workforce transition has been a significant issue, with drivers retrained for roles in control rooms and maintenance.

# Autonomous haulage: compare productivity of staffed and autonomous truck fleets
import numpy as np
rng = np.random.default_rng(21)
payload_t, cycle_min = 320, 35.0                # tonnes per load, average cycle time
hours_year = 8760
def annual_tonnes(utilization, n_trucks, cycle_sd):
    cycles = rng.normal(cycle_min, cycle_sd, size=(n_trucks, 2000)).clip(25, 60)
    loads_per_hour = 60 / cycles.mean()
    return n_trucks * loads_per_hour * payload_t * hours_year * utilization
staffed = annual_tonnes(utilization=0.70, n_trucks=20, cycle_sd=5.0)    # shift changes, breaks
auton = annual_tonnes(utilization=0.80, n_trucks=20, cycle_sd=3.0)      # fewer interruptions
print(f"staffed fleet:     {staffed / 1e6:6.1f} Mt per year")
print(f"autonomous fleet:  {auton / 1e6:6.1f} Mt per year  (+{100 * (auton / staffed - 1):.0f} %)")

staffed fleet:       67.2 Mt per year
autonomous fleet:    76.9 Mt per year  (+14 %)

Agriculture

Farms have used automation for decades. GNSS auto-steer, using RTK corrections from Lesson 3, guides tractors along parallel passes with centimetre accuracy and is now standard on large farms across the Canadian Prairies and the US Midwest, reducing overlap, fuel and operator fatigue. The next step is driverless operation: equipment makers have introduced autonomous tractors for tillage and other tasks that a farmer supervises remotely from a phone or tablet, using cameras and machine learning to detect obstacles. Smaller autonomous robots target weeding, scouting and specialty crops, and agricultural drones spray and map fields under the drone rules of Lesson 7.

Agriculture suits autonomy because fields are private, sparsely populated and structured, and because labour is scarce at critical times such as seeding and harvest. Challenges include connectivity in rural areas, detecting people or animals in tall crops, the variety of implements and soil conditions, the cost of equipment for smaller farms, and questions about who owns and controls farm data. Farmer right-to-repair concerns also apply to autonomous equipment.

Sidewalk delivery robots

Small wheeled robots that travel on sidewalks at walking speed deliver food and parcels on university campuses and in some urban neighbourhoods. They are slow, light, typically under about 50 kg loaded, and can stop quickly, so the kinetic energy involved is small, and they are often supervised remotely by people who handle difficult crossings, with one supervisor monitoring several robots. Many US states have passed laws defining personal delivery devices and setting weight and speed limits for them. In Canada, rules are provincial and municipal: Ontario created a pilot framework that lets municipalities allow such devices under conditions, while Toronto chose to prohibit them on its sidewalks. The debate centres on sharing scarce sidewalk space, especially with people using wheelchairs, walkers or white canes, for whom a robot blocking a curb ramp is a real barrier. Snow-covered sidewalks pose an additional challenge in Canadian winters.

Lessons across domains

Comparing these domains suggests a pattern. Autonomy is easiest where the environment is controlled, the operator controls the whole system, interactions with the public are few, speeds or energies are modest, and the economic value of each automated hour is high. Rail metros and mines satisfy most of these and automated early; agriculture and highway trucking satisfy several and are now deploying; urban robotaxis and sidewalk robots face open environments and the public and require the most validation. For each application, defining a narrow, well-understood operational design domain and expanding it gradually, as in Lesson 1, is the common thread.