Autonomous Trucks Are Here To Stay and Here’s One Reason
- 4 hours ago
- 13 min read
What the federal record actually proves about driver distraction—and what Volvo’s one-in-ten forecast could mean through 2035
By Micheal “Cupcake” Cobb
Primary-source analysis using NHTSA, FARS, FMCSA, NTSB, eCFR, FHWA, and Volvo Group records.
Let’s start with the part nobody needs a Ph.D. to understand: a truck with no driver cannot have a driver watching TikTok, streaming a movie, texting, scrolling, eating a three-course dashboard dinner, or arguing with dispatch while rolling down the highway. Remove the in-cab driver from the driving task and you remove in-cab driver distraction from that task. Period.
That is a real safety mechanism. It is not, by itself, proof that an autonomous truck is safer overall. There are legitimate arguments about autonomous trucks, and there is also a mountain of bullshit coming from both sides. So this article separates what the federal record proves from what people merely want it to prove—and then runs Volvo’s one-in-ten buildout without dressing a company forecast up as a fact. [11]
THE DEFENSIBLE CLAIM
True Level 4 driverless operation eliminates the in-cab human-driver distraction pathway inside its operational design domain. Federal data do not yet measure the net crash reduction from that change or prove lower all-cause risk. [8]
The federal record says distraction is real—and badly measured
NHTSA’s latest national report estimated that distraction affected 732,914 of 6,180,241 police-reported crashes in 2024—11.86%. It counted 2,955 distraction-affected fatal crashes, 3,208 deaths, and an estimated 315,167 injured people. Within the fatal-crash data, 404 crashes and 437 deaths were associated with cellphone use. Those are not truck-only numbers; they cover the national road system. [1]
Now for the part that ruins every too-clean talking point: NHTSA says distraction is difficult to identify, is likely underreported, varies with police-report fields and investigative practices, and may be coded as present without being the sole cause of the crash. “Distraction-affected” means involvement in the record. It does not automatically mean “this crash disappears if distraction disappears.” [1]
Federal CMV rules already prohibit a driver from texting while driving and prohibit use of a hand-held mobile telephone while driving, with limited emergency exceptions. The rule exists because the hazard is not theoretical. The compliance problem is painfully human: a prohibition cannot make a person keep both eyes and both hands where they belong every second of every trip. [6]
Commercial-truck research is not subtle
FMCSA’s landmark naturalistic CMV study followed 203 drivers in 55 trucks across seven fleets, covering roughly 3 million miles. It captured 4,452 safety-critical events and 19,888 ordinary-driving baselines. Texting was associated with 23.24 times the odds of a safety-critical event, with a 95% confidence interval of 9.69 to 55.73. During a six-second texting task, drivers’ eyes were away from the forward road for an average of 4.6 seconds. Dispatching, writing, calculator use, map use, dialing, reading, and other electronic-device tasks also carried elevated odds. [4]
A later FMCSA analysis covered more than 3.8 million miles from 182 trucks and 172 drivers. Browsing and texting increased safety-critical-event risk, and glances away from the roadway longer than two seconds were statistically significant. That study used data collected from 2012 through 2015 and mostly local or regional operations, so it should not be sold as a perfect national portrait. It is still strong evidence for the mechanism: taking a driver’s attention off the road increases risk. [5]
One more guardrail: an odds ratio is not “the percentage of crashes caused by texting.” Twenty-three times the odds within one study does not mean 23 times as many crashes across the whole country. Anyone telling you otherwise is doing math with a chainsaw.
What the 2024 tractor records show
I ran a reproducible cut of NHTSA’s 2024 FARS National CSV. The selection used truck tractors—BODY_TYP 66—and actual driver records—PER_TYP 1—joined by state, crash number, and vehicle number. Because FARS only contains qualifying fatal crashes, every driver in this cut was already connected to a fatal crash. This is a conditional fatal-crash sample, not a national distraction-prevalence survey. [2]
The cut produced 2,842 tractor-driver records. Of those, 126 were coded as distracted, or 4.43%, and 29 were phone-coded, or 1.02%. Those 126 records appeared in 125 unique fatal crashes with 146 total deaths: 31 in the subject tractors, 95 occupants of other vehicles, and 20 nonoccupants. The 29 phone-coded crashes contained 32 deaths. [2]
But 1,381 records were coded “not reported” and 221 were “unknown.” Together, that is 56.4% of the tractor-driver denominator. The reported 4.43% is therefore not the true prevalence of distraction among fatal-crash-involved tractor drivers. It is the share that was affirmatively coded in this dataset. That distinction is not a small footnote. That is the whole ballgame. [2]
FARS also cannot give us a national “TikTok crashes” or “movie-watching truck crashes” number. The codes include categories such as manipulating a phone, talking or listening, other phone-related use, reaching for a device, and distraction or inattention with details unknown. Inventing a TikTok-specific national number would be fake precision wearing a lab coat. [2]
Does removing distraction remove every distraction-coded crash?
No—not automatically. It removes every crash for which human distraction was a necessary link in the chain. If the distracted driver would have braked, steered, or recognized the hazard in time while attentive, removing that distraction can remove the crash. If a tire fails, another vehicle crosses the median, a load shifts, black ice wins, or the crash would have occurred anyway, the presence of a distraction code does not prove preventability.
For that reason, this article treats 100% prevention of selected distraction-involved events as an upper-bound scenario. It answers, “What is the ceiling if we grant the strongest version of the argument?” It does not answer, “What will happen?” That would require causal case review or matched real-world crash rates from deployed driverless trucks.
THE CAUSAL RULE
No driver means no driver distraction. But “distraction was coded” and “distraction caused the crash” are not interchangeable statements. [1] [2]
True driverless means Level 4—not a human babysitting the machine
FMCSA’s federal-use definition of SAE Level 4 is straightforward: inside a defined operational design domain, the automated driving system performs the entire dynamic driving task and the fallback without expecting a user to respond to a request to intervene. That is the category relevant to the “no distracted driver” claim. Level 2 and Level 3 systems, developmental systems with a safety operator, manual operation, and operation outside the Level 4 domain belong in separate buckets. [8]
The NTSB’s Tempe investigation shows why that separation matters. A developmental ADS was operating a passenger SUV while a human operator was supposed to supervise it. Phone records showed a television show streaming throughout the 39-minute road trip. The operator looked down toward the phone area for 34% of the moving trip and for five of the final six seconds before impact. A pedestrian was killed. NTSB identified the operator’s visual distraction along with inadequate safety-risk management and automation complacency. [7]
That crash was not a Class 8 truck and was not true no-fallback Level 4 service. It does not prove autonomous trucks are unsafe. It proves something narrower and crucial: supervised automation can preserve the distraction problem—and may invite complacency—when the human is still the fallback. You cannot credit “driver removal” to a system that still depends on a human paying attention. [7] [8]
Volvo’s one-in-ten forecast, built out year by year
Volvo’s June 2026 Capital Markets Day presentation says: “In less than ten years, one in ten heavy-duty long-haul trucks on U.S. highways will be driverless.” The preceding slide forecasts a 25,000-truck installed base in 2030 and 220,000 in 2035. The roadmap then puts driverless launch in Q1 2027 and more than 300 autonomous trucks in operation by the end of that year. So the sourced endpoint for this article is 220,000—not a round number borrowed from somewhere else. Volvo’s category is also narrower than every combination truck registered in America. [11]
The ten-year buildout below covers 2026 through 2035. It starts at zero in 2026 because Volvo schedules driverless launch for 2027, uses 300 exactly as a conservative floor for Volvo’s “more than 300” year-end milestone, then passes through Volvo’s 25,000 and 220,000 installed-base points. The intervening years are straight-line interpolation. Those annual in-between values are this model’s math—not Volvo promises. [11]
The mileage premise remains 250,000 miles per autonomous truck per year: the requested 125,000-mile human baseline doubled to reflect Volvo’s stated two-times asset-utilization claim. Volvo did not publish the 250,000-mile figure. At 50 mph, each modeled truck logs 5,000 moving hours annually—about 13.7 moving hours and 685 miles per calendar day. In 2035, 220,000 trucks therefore produce 55 billion miles and 66 billion moving minutes. Aggressive? Yes. Literally nonstop? No. Maintenance, inspections, loading, fuel or charging, weather, blocked routes, and operational-domain limits still stop the truck. [11]
Volvo’s 220,000 endpoint and one-in-ten statement imply a 2.2-million-truck heavy-duty long-haul base. If the other 90% run 125,000 miles and the driverless 10% run 250,000, that 10% of units performs 18.18% of the modeled fleet miles—the mileage of 440,000 human-utilization trucks. Unit share alone can therefore understate the exposure and the safety stakes. [11]
Across the full 2026–2035 buildout, the interpolated curve totals 760,600 autonomous truck-years, 190.15 billion miles, and 228.18 billion moving minutes. [11]
What the crash math produces
NHTSA reports 548,521 large trucks involved in police-reported crashes during 2024 over 329.6 billion large-truck miles. That works out to about 166.42 large-truck crash involvements per 100 million miles. “Involvement” matters: two large trucks in one crash can create two truck involvements. Applying that national all-severity rate to the scenario supplies a baseline exposure count—not a prediction tailored to a driverless operational domain. [3]
Year
Driverless trucks
Share of 2.2M base
Annual miles
Moving minutes
Upper-bound potentially avoided crash involvements
2026
0
0.00%
0
0
0
2027*
300
0.01%
75M
90M
6–15
2028
8,533
0.39%
2.13B
2.56B
157–421
2029
16,767
0.76%
4.19B
5.03B
309–827
2030*
25,000
1.14%
6.25B
7.5B
461–1,233
2031
64,000
2.91%
16B
19.2B
1,181–3,158
2032
103,000
4.68%
25.75B
30.9B
1,900–5,082
2033
142,000
6.45%
35.5B
42.6B
2,619–7,006
2034
181,000
8.23%
45.25B
54.3B
3,339–8,930
2035*
220,000
10.00%
55B
66B
4,058–10,855
* Volvo-published milestones. The model uses 300 trucks for Volvo’s “more than 300” 2027 milestone, sets 2026 to zero because launch is scheduled for Q1 2027, and linearly interpolates unstarred years between Volvo’s 2027, 2030, and 2035 points. Each truck is assigned 250,000 miles a year at 50 mph. The avoidance range applies the 4.43% and 11.86% proxies as 100% prevention ceilings—not expected outcomes. [1] [2] [3] [11]
In 2035, the model produces roughly 91,531 baseline large-truck crash involvements over 55 billion miles. Applying the 4.43% FARS tractor-driver share yields an upper-bound 4,058 potentially avoided involvements. Applying the 11.86% national all-vehicle share yields 10,855. Across the full ten-year buildout, the baseline is about 316,448 involvements and the two upper-bound avoidance figures are 14,030 and 37,528. [1] [2] [3] [11]
That 4,058-to-10,855 spread is not a confidence interval and should not be presented as “the answer.” The lower proxy comes from tractor-driver records already inside fatal crashes; the upper proxy comes from all police-reported crashes across all vehicle types. Applying either one to all-severity large-truck involvements mixes categories. The exercise is useful for scale and sensitivity. It is not a statistically validated forecast. [1] [2] [3]
The FARS fatal-crash line is a second rough ceiling. Scaling 125 distraction-coded tractor fatal crashes and 146 deaths to the scenario’s mileage gives about 35.7 fatal crashes and 41.7 deaths in the 2035 exposure, and about 123.5 crashes and 144.2 deaths across the buildout. Again: “potentially avoided if every selected case required human distraction,” not “expected lives saved.” [2] [9] [11]
What hours of service do—and do not—change
Hours-of-service rules in 49 CFR Part 395 govern motor carriers and drivers. A true driverless Level 4 truck has no in-cab human driver whose personal 11-hour driving limit forces the vehicle to stop. That is a practical inference from a driver-based framework, not a magic exemption from every federal safety duty. Any human who still drives or performs a regulated role remains subject to applicable rules, and the carrier, vehicle, maintenance, inspection, and operational limits do not evaporate. [8] [10]
So, yes: higher utilization is a legitimate autonomous-truck premise. No: “they never have to stop” is not. The 250,000-mile input already leaves substantial nonmoving time, and that is the honest way to model it.
Removing one human failure does not grant the software sainthood
A truly driverless truck cannot get bored, sleepy, drunk, angry, or sucked into a video. It also cannot use human common sense when the lane markings disappear, a trooper hand-signals traffic, a work zone contradicts the map, a sensor is blinded, cargo shifts, or a damaged truck needs to be secured. Those risks do not cancel the distraction benefit. They sit on the other side of the ledger.
The right comparison is therefore not “distracted human versus flawless robot.” It is matched, all-cause harm per mile: true driverless Level 4 operation against comparable human-driven operation on the same roads, in the same weather, at the same times, with similar loads and crash-reporting rules. Removing one dumb human failure mode is valuable. Pretending it removes every failure mode is salesmanship.
The federal data package needed to prove net safety
A credible national evaluation should require the following before anyone claims victory:
· Driverless-only miles, separated from safety-driver, remote-driver, manual, and supervised-automation miles.
· A matched human-driven control group by road class, geography, speed, weather, time, cargo, and operational design domain.
· Common definitions for crashes, injury severity, tow-away events, roadway departures, disengagements, minimal-risk stops, and safety-critical events.
· All-cause rates per mile and per operating hour—not raw counts and not company-selected anecdotes.
· System version, sensor condition, maintenance history, remote-support involvement, and whether the truck was inside its approved domain.
· Independent data audit, public aggregate reporting, and consequences for late, incomplete, or misleading submissions.
· Separate accounting for people outside the truck, because large-truck crashes impose most fatal harm on other road users. [3]
Until that package exists at meaningful scale, the honest conclusion is asymmetric: the mechanism is proven, the magnitude is not. We know removing the in-cab driver removes in-cab driver distraction. We do not yet know the net national crash-rate change after system, maintenance, remote-support, infrastructure, and operational-domain failures are counted.
Bottom line
My original point survives. In fact, it gets stronger when it is stated precisely instead of exaggerated: true Level 4 driverless trucks remove the possibility that an in-cab driver is watching TikTok, streaming a movie, texting, or otherwise failing to watch the road. Federal CMV research shows that those behaviors and long off-road glances sharply increase safety-critical-event risk. [4] [5] [8]
What does not survive is the leap from “the distraction pathway is gone” to “every distraction-coded crash is gone” or “autonomous trucks are therefore safer overall.” Volvo’s one-in-ten endpoint shows the possible scale: in 2035, 55 billion miles and 66 billion moving minutes, with a deliberately aggressive ceiling of roughly 4,058 to 10,855 all-severity truck crash involvements tied to the two distraction proxies. Across the ten-year buildup, the comparable ceiling is about 14,030 to 37,528. That is a scenario, not a promise. [1] [2] [3] [11]
THE SENTENCE I WOULD PUBLISH
True Level 4 driverless trucks eliminate in-cab driver distraction because no in-cab human is performing—or expected to resume—the driving task within the operational design domain. That can eliminate crashes that genuinely require human distraction, but current federal data do not quantify how many or prove lower net crash risk. [1] [2] [8]
That is not walking the statement back. That is making it bulletproof.
Methodology note
The projection is an exposure scenario, not a statistical correlation from paired observations. Volvo supplies three adoption anchors: more than 300 trucks by the end of 2027, 25,000 in 2030, and 220,000 in 2035. The model sets 2026 to zero, uses 300 in 2027, and linearly interpolates every unreported year between the published anchors. Annual miles equal modeled trucks × 250,000; the 250,000-mile assumption doubles the requested 125,000-mile human baseline to reflect Volvo’s two-times asset-utilization claim. Moving minutes equal miles ÷ 50 mph × 60. Baseline crash involvements equal projected miles × (548,521 ÷ 329.6 billion). The two all-severity avoidance scenarios multiply that baseline by 126 ÷ 2,842 and 732,914 ÷ 6,180,241. The fatal-case ceiling scales 125 selected FARS crashes and 146 deaths by projected miles ÷ 192.52 billion FHWA combination-truck miles. Each cross-application has category and denominator limitations described in the article. [1] [2] [3] [9] [11]
The custom FARS cut used vehicle.csv BODY_TYP = 66 and person.csv PER_TYP = 1, joined to distract.csv and accident.csv by STATE, ST_CASE, and VEH_NO. Phone-coded records were FARS codes 5, 6, and 15. Counts were checked at the driver, unique-crash, and fatality levels. No case-level causal review was performed. [2]
Primary sources
Every factual citation in this article points to a primary source. Federal records supply the safety, crash, regulatory, and exposure evidence; Volvo’s own presentation supplies only Volvo’s adoption forecast and utilization claim. The short notes below say what each source can support—and where the guardrail belongs.
[1] National Highway Traffic Safety Administration, National Center for Statistics and Analysis. Distracted Driving in 2024. Traffic Safety Facts Research Note, DOT HS 813 790. April 2026.
Use / limitation: National fatal-crash, fatality, injury, and police-reported crash estimates. NHTSA warns that distraction is underreported and that a coded distraction-affected crash does not by itself establish causation.
[2] National Highway Traffic Safety Administration, National Center for Statistics and Analysis. Fatality Analysis Reporting System 2024 National CSV, Annual Report File. FARS 2024 raw data. 2026 release.
Use / limitation: Raw source for the custom truck-tractor fatal-crash analysis. FARS is a census of qualifying fatal crashes, not a measure of all crashes or total driving exposure.
[3] National Highway Traffic Safety Administration, National Center for Statistics and Analysis. Traffic Safety Facts 2024 Data: Large Trucks. DOT HS 813 816. 2026.
Use / limitation: Official 2024 large-truck crash involvements and vehicle-miles traveled. Counts are truck involvements, not necessarily unique crash events.
[4] Federal Motor Carrier Safety Administration. Driver Distraction in Commercial Vehicle Operations. FMCSA-RRR-09-042. September 2009.
Use / limitation: Naturalistic CMV study of task-related safety-critical-event risk. Odds ratios describe association within the study; they are not percentages of crashes caused.
[5] Federal Motor Carrier Safety Administration. Analysis of Naturalistic Driving Data to Assess Distraction and Drowsiness in Drivers of Commercial Motor Vehicles. FMCSA-RRR-20-003. August 2021.
Use / limitation: Later naturalistic CMV evidence on browsing, texting, and prolonged off-road glances. The fleets were not a random national sample and data were collected from 2012 through 2015.
[6] Electronic Code of Federal Regulations / Federal Motor Carrier Safety Administration. 49 CFR Part 392, Subpart H — Limiting the Use of Electronic Devices. 49 CFR 392.80 and 392.82. current through 27 August 2026.
Direct URL: https://www.ecfr.gov/current/title-49/subtitle-B/chapter-III/subchapter-B/part-392/subpart-H
Use / limitation: Current federal prohibitions on texting and hand-held mobile-phone use by CMV drivers, subject to stated exceptions.
[7] National Transportation Safety Board. Collision Between Vehicle Controlled by Developmental Automated Driving System and Pedestrian, Tempe, Arizona, March 18, 2018. Highway Accident Report NTSB/HAR-19/03, PB2019-101402. adopted 19 November 2019.
Use / limitation: Investigated example of distraction during human supervision of a developmental ADS. It involved a passenger SUV, not a Class 8 truck or a true no-fallback driverless operation.
[8] Federal Motor Carrier Safety Administration. Safe Integration of Automated Driving Systems-Equipped Commercial Motor Vehicles. Federal ADS-CMV policy and definition context. 22 May 2019.
Use / limitation: Federal-use definitions of SAE Level 4 and Level 5 and discussion of unmanned CMV operation inside an operational design domain. A definition is not proof of safety performance.
[9] Federal Highway Administration, Office of Highway Policy Information. Highway Statistics 2024, Table VM-1 — Annual Vehicle Distance Traveled in Miles and Related Data. Table VM-1. updated 2 February 2026.
Use / limitation: Current combination-truck registrations and vehicle-miles traveled used as national exposure context and as the denominator for the fatal-case sensitivity line. It is not used to convert Volvo’s 10% long-haul forecast into a share of every registered combination truck.
[10] Electronic Code of Federal Regulations / Federal Motor Carrier Safety Administration. 49 CFR Part 395 — Hours of Service of Drivers. 49 CFR Part 395. current through 27 August 2026.
Use / limitation: Current driver hours-of-service framework. The inference about an absent driver having no personal driving-hour clock does not eliminate carrier, vehicle, maintenance, inspection, ODD, or any human-operator obligations.
[11] AB Volvo. Volvo Group Capital Markets Day 2026 — Presentation Material. Volvo Autonomous Solutions, slides 59–63. 10 June 2026.
Use / limitation: Volvo’s company forecast and roadmap: more than 300 autonomous trucks operating by the end of 2027, a 25,000-truck installed base in 2030, a 220,000-truck installed base in 2035, one in ten U.S. heavy-duty long-haul highway trucks driverless within the decade, and two-times asset utilization. These are Volvo forecasts and business claims, not federal projections or demonstrated safety outcomes.
Comments