BRENNEN

Industries / Energy, water and waste / Fuel jobbers and wholesale fuel

Trucks and dispatch

Trucks and dispatch is the work of loading tank trucks at the terminal, deciding which truck takes which load, and getting the right fuel into the right customer tank safely.

7

Milestones achieved, from 1910 to 2018.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1910

    Motor tank trucks arrive

    Standard Oil's European subsidiary Anglo-American may have been first with motor tank trucks, using Thornycroft trucks from 1905 to carry oil from railroad stations to substations, and tank trucks began to appear in the United States in 1910 and 1911, by petroleum historian Samuel T. Pees's account. They took over from horse-drawn tank wagons of 300 to 900 gallons, which already split a load into compartments: a Standard Oil wagon of about 1892, now at The Henry Ford museum, has three, for kerosene, lubricating oil and gasoline. The Henry Ford says horse-drawn wagons were still the main way to move these products from the depot to the customer as late as the 1920s.

    petroleumhistory.orgpetroleumhistory.orgthehenryford.org

    Credit

    Standard Oil companies, including Anglo-American and Conoco; Morrison Brothers Company, which built the 1892 wagon

  2. 1981

    Computers dispatch the trucks

    Gerald Brown and Glenn Graves described a real-time dispatch system in daily use at Chevron U.S.A. It assigned about 2,600 loads a day from more than 80 bulk terminals to a fleet of over 300 tank trucks, in place of extensive manual work, while keeping workloads even and matching each load to equipment able to carry it. By their count the fleet's transport costs approached $100 million a year. In 1987 Brown, Graves and two co-authors reported a similar system at Mobil Oil: 120 bulk terminals, more than 430 vehicles, 600,000 customer orders a year, and about one second of computer optimization per dispatch.

    doi.orgdoi.org

    Credit

    Gerald G. Brown and Glenn W. Graves; Chevron U.S.A.; Mobil Oil Corporation

  3. 2005

    Background checks for hazmat drivers

    A US truck driver needs a tank vehicle endorsement on the commercial driver's license (CDL) to drive a tank truck, and a hazardous materials endorsement to haul loads such as gasoline. The USA PATRIOT Act of 25 October 2001 barred states from issuing that hazmat endorsement until the federal government had found the driver not to be a security risk. The Transportation Security Administration (TSA) set the standards in a May 2003 rule, counted about 3.5 million endorsement holders, and moved fingerprint-based background checks in every state to 31 January 2005.

    law.cornell.edufederalregister.govfederalregister.gov

    Credit

    US Congress; Transportation Security Administration

  4. 2015

    A wetlines rule withdrawn

    Wetlines are the loading pipes under a tank truck. After loading they can hold up to 50 gallons of fuel, which can make a crash with a car worse. The National Transportation Safety Board (NTSB), in recommendation H-98-27, asked for hazardous materials to be kept out of such pipes, and US regulators proposed rules in 2004 and 2011. The Pipeline and Hazardous Materials Safety Administration (PHMSA) counted 132 wetline incidents from January 1999 to March 2011 and put the best-case ratio of benefits to costs at 0.78. In 2013 the Government Accountability Office (GAO) found that PHMSA's incident form had no box for a wetline incident, so the data could not show how many there were. The FAST Act of 4 December 2015 ordered the proposal withdrawn, and PHMSA withdrew it on 30 December 2015.

    Lessonwithout incident records that name the hazard, the safety case could be neither made nor refuted.

    federalregister.gov

    Credit

    NTSB; PHMSA; GAO; US Congress

  5. 2017

    Stability control on tractors

    US safety standard FMVSS No. 136 requires electronic stability control on typical three-axle truck tractors built from 1 August 2017. The system brakes single wheels, including trailer brakes, and cuts engine power when sensors show a rollover or skid starting. The agency that wrote it, NHTSA, notes that fuel sloshing in a partly full tanker can sharply reduce roll stability, and by its own estimate the rule will prevent 40 to 56% of rollovers not caused by hitting something. The NTSB had asked for stability systems on cargo tank trailers already in service; the rule covers new tractors only.

    federalregister.gov

    Credit

    National Highway Traffic Safety Administration (NHTSA)

  6. 2017

    Crowds at an overturned tanker

    On 25 June 2017 a tanker of petrol overturned near Ahmedpur East in Bahawalpur district, Pakistan. Hundreds of villagers gathered to collect the leaking fuel, and the tanker exploded in a fireball; Dawn put the death toll at 214. Pakistan's Oil and Gas Regulatory Authority (OGRA) found the truck failed its 2009 technical standards for carrying petrol and had a fake fitness certificate. The truck belonged to a contractor, but OGRA held Shell Pakistan responsible as its licensee, fined it Rs10 million and estimated that up to 85% of the country's oil tank lorries did not meet the standards.

    Lessonthe fuel company that hires a haulier answers to the regulator for the haulier's truck.

    dawn.comdawn.com

    Credit

    OGRA, in its inquiry

  7. 2018

    Diesel delivered to the door

    In March 2018 Indian Oil Corporation began a trial of doorstep diesel delivery in Pune: a dispenser like a station pump, mounted on a midsized truck with a storage tank, for customers such as malls running generators, transport firms and farm harvesters. Its chairman, Sanjiv Singh, said it was the first company to start after clearance from India's safety regulator, the Petroleum and Explosives Safety Organisation (PESO). PESO had approved diesel only, on a trial basis, and also cleared trials for Hindustan Petroleum and Bharat Petroleum. A Bengaluru startup, ANB Fuels, had started a similar service in June 2017 and was stopped by PESO.

    Lessondelivery from a truck-mounted pump came only after the safety regulator approved the equipment.

    business-standard.combusinesstoday.in

    Credit

    Indian Oil Corporation; PESO

Watch for

  1. Software builds the loads

    Gravitate, a US software company, says its AI-assisted dispatch system builds a load in 90 seconds, against 44 minutes for the best human dispatcher, at CLI Transport, the carrier that delivers all fuel to Sheetz's more than 800 US stores. Eamonn Clarke of CLI told a webinar of 40 to 50% gains in dispatcher efficiency, no new dispatcher hires needed for two years or more, and retains (fuel left on the truck after a run) close to zero. Gravitate's case study says more than 200,000 loads have run since the cutover, with no store running out.

    Achieved whenan outside party compares dispatch time, run-outs, retains and cost per gallon before and after at Sheetz or another named fleet.

    gravitate.energywelcome2.studygroups.comgravitate.energy

    Claimed by

    Gravitate and CLI Transport, October 2025 and January 2026

Open

  1. 01

    Tag tank-truck spills by cause

    A public table of every US incident report since 2010 for a cargo tank truck carrying gasoline (UN1203) or diesel, with each report tagged by cause from its written description: a struck wetline, an overfilled customer tank, fuel put in the wrong tank, a hose or valve failure while unloading, or a rollover. An AI model reads each description, and people check a sample.

    The brief
    Credit · unclaimed
  2. 02

    Fuel loads on real maps

    A public set of fuel delivery problems built from real places: station locations from OpenStreetMap, drive times from an open road router, several terminals to load from, trucks with four to six compartments, grades that may not share a compartment, the empty space in each station's tanks, and drivers' shift limits. Include a checker and a results table for free solvers and for AI agents acting as dispatchers, with the time each takes.

    The brief
    Credit · unclaimed

Open 01 · the brief

Tag tank-truck spills by cause

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public table of every US incident report since 2010 for a cargo tank truck carrying gasoline (UN1203) or diesel, with each report tagged by cause from its written description: a struck wetline, an overfilled customer tank, fuel put in the wrong tank, a hose or valve failure while unloading, or a rollover. An AI model reads each description, and people check a sample.

Why it is still open

In 2013 the GAO found that the federal incident form has no option for a wetline incident, so such incidents cannot be counted from the data. PHMSA's own review cut 172 candidate incidents to 132 by hand, and the 2015 withdrawal still turned on whether three or four of them were fatal. In September 2026 PHMSA proposed a revised incident form; its notice says many written descriptions run to only a sentence or two, and it codes each package failure from a fixed list of what failed, how and why. This session found studies that link these reports to crash records, but no public tagging of the written descriptions by cause.

federalregister.govpublic-inspection.federalregister.gov

Built with what exists

PHMSA's Hazmat Incident Report search tool, which exports reports to text files; an AI model that returns probabilities for fixed-answer questions (Note: "A thousand questions, one pass"); Python and pandas.

catalog.data.gov

  1. First step

    Export one year of cargo tank reports for gasoline and tag 50 descriptions by hand to settle the categories.

  2. Done looks like

    All qualifying reports tagged, each row linked to its report number; a random sample of 300 tagged separately by two people, with the model agreeing with them on at least 90%; yearly counts by cause published.

Open 02 · the brief

Fuel loads on real maps

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public set of fuel delivery problems built from real places: station locations from OpenStreetMap, drive times from an open road router, several terminals to load from, trucks with four to six compartments, grades that may not share a compartment, the empty space in each station's tanks, and drivers' shift limits. Include a checker and a results table for free solvers and for AI agents acting as dispatchers, with the time each takes.

Why it is still open

Operations researchers call this the petrol station replenishment problem and have studied it for decades. A 2024 GERAD report calls it a problem "with many variants in the literature", and studies test methods on their own case data, such as 59 stations over 14 days at a Danish distributor. Public test sets cover parts of the problem on made-up maps: Leandro Coelho hosts the 670 multi-compartment delivery problems that he and Gilbert Laporte published in 2015, with several products, compartments and customer storage limits, but customers placed at grid coordinates; an ETH Zurich student project with the Swiss firm Arqh released code in 2025, under the MIT license, that generates random replenishment problems with one product and no compartments. Vendor claims such as Gravitate's 90 seconds a load come without public test problems. This session found no public set on real road networks with a choice of terminals, and no published scores for AI agents on such problems.

gerad.carepository.tudelft.nlleandro-coelho.comgithub.com

Built with what exists

PyVRP or Google OR-Tools for routes; HiGHS for load assignment; OpenStreetMap and OSRM, an open road router; files in the CVRPLIB format used for routing tests; two or more AI models run as dispatch agents.

  1. First step

    Build one problem by hand from 20 real station locations in one county, three grades and five-compartment trucks, and solve it with OR-Tools.

  2. Done looks like

    At least 100 problems in three sizes from at least five regions; a checker that rejects any plan that mixes grades in a compartment, overfills a tank or breaks a shift limit; published results for at least three solvers and two AI agents, with the same solvers also run on Coelho and Laporte's set as a cross-check.