Industries / Energy, water and waste / Propane distribution
Delivery
This is the work of getting propane from the bulk plant to each customer, either by small tank trucks called bobtails that pump it into the customer's own tank or by swapping full cylinders for empty ones, and of deciding which customers to visit on which day.
5
Milestones achieved, from 1999 to 2020.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1999
A remote stop for bobtails
On 8 September 1996 a hose came off its coupling while a transport truck unloaded at a propane bulk plant in Sanford, North Carolina, and more than 35,000 gallons escaped; it did not ignite. The truck's excess-flow valve, meant to shut by itself when a hose breaks, never closed. Tests by Mississippi Tank Company then showed that the pump fitted to speed deliveries holds the flow down, so the valve never trips, and regulators counted nine similar releases from small delivery trucks in the previous ten years. A final rule of May 1999 defined "metered delivery service" (100 gallons a minute or less through a hose 1.25 inches wide or narrower), told the driver to stay within 150 feet of the truck and 25 feet of the hose and to look at the truck, the customer's tank and the hose at least every five minutes, and required a remote control, carried by the driver, that closes the truck's internal valve and stops its engine and pump and cannot reopen the valve. By the rule's own figures the average such delivery is under 170 gallons, pumped at 40 to 60 gallons a minute through a hose 100 to 150 feet long.
Lessona valve meant to close on a broken hose stopped working once pumps were added, so the means to stop the flow was put in the driver's hand.
Credit
the US Research and Special Programs Administration and Federal Highway Administration, with a task force of the National Propane Gas Association
2008
Exchange cylinders hold less
Cylinder exchange lets a grill owner trade an empty cylinder for a filled one at a shop. By the US Federal Trade Commission's account, Blue Rhino (part of Ferrellgas) and AmeriGas Cylinder Exchange together sold about 80% of the wholesale exchange cylinders in the United States, and in 2008 each cut the propane in its cylinders from 17 to 15 pounds without lowering the wholesale price. The FTC's complaint of 27 March 2014 alleged that when Walmart refused the cut, the two companies agreed that neither would back down. Both agreed to settle on 31 October 2014, under orders that bar them from agreeing with competitors on fill levels or on the prices of exchange cylinders.
Lessonin cylinder exchange the amount of gas in the cylinder is part of the price, and regulators treat an agreement on it as an agreement on price.
Credit
Ferrellgas (Blue Rhino) and UGI's AmeriGas; case brought by the Federal Trade Commission
2009
Routes planned by software
In March 2009 the trade magazine CCJ reported that AmeriGas, the largest US propane retailer, was extending hosted routing, dispatch and vehicle-tracking software from Descartes Systems Group across a segment of its national delivery operation of more than 3,000 vehicles, so that dispatchers could change routes during the day. An undated Descartes case study of Ferrellgas's "Project Renaissance" says Ferrellgas, with about 2,600 delivery trucks each making 20 to 25 stops a day, folded 600 local offices into 150 service centers, put route software on every truck's onboard computer and cut its fleet requirement by 11%.
Credit
AmeriGas and Ferrellgas, with Descartes Systems Group
2014
A propane-fuelled bobtail falters
In April 2014 Freightliner Custom Chassis handed AmeriGas the first field-test unit of its S2G, a work truck built to run on propane, the fuel a bobtail carries. The first 25 units were to go through a delivery trial before production began in mid-2014. In October 2025 LP Gas magazine described the S2G as an attempt that did not "result in a consistently successful product", and said a Cummins B6.7 propane engine never reached the market; the Propane Education & Research Council (PERC), the industry's research and training fund, had approved millions of dollars for both.
Lessonrunning delivery trucks on propane has taken more than a working engine; retailers commenting on the next attempt in 2025 named the truck's design and its service network as their doubts.
Credit
Freightliner Custom Chassis and AmeriGas, with funding from PERC
2020
A code at the door
From 1 November 2020 India's oil marketing companies began completing home deliveries of LPG cylinders only after the customer gives the delivery worker a delivery authentication code (DAC), a one-time code sent to the customer's registered mobile phone at booking. It started in 100 "smart cities" after a pilot in Jaipur, to stop cylinders being stolen or handed to the wrong person. In March 2026, during a supply crisis tied to the conflict in West Asia, the Free Press Journal reported that the code covered about 72% of deliveries, up from 53%, and that the Ministry of Petroleum and Natural Gas had told the oil companies to reach at least 80%.
Credit
India's oil marketing companies and the Ministry of Petroleum and Natural Gas
Watch for
A bobtail that burns propane
Nexio Power says its production-intent Class 7 bobtail, with a 7.2-liter supercharged V8 engine that runs on propane, went into commercial service on daily routes in South Texas with PICO Propane and Fuels in August 2026. PICO is to share route data to support certification by the US Environmental Protection Agency (EPA) and the California Air Resources Board (CARB), backed by $6 million that PERC approved in July 2025. Nexio claims $25,000 a year in fuel savings over a diesel truck.
Achieved whenthe engine holds EPA and CARB certification and a retailer other than Nexio publishes at least one winter of route data for the truck, such as fuel cost per gallon delivered and days out of service.
Claimed by
Nexio Power, August 2026
Open
01
Score forecasts on real tanks
A public set of daily tank readings from at least 100 homes that already have remote tank monitors, donated with the owners' consent together with tank size, a coarse location and delivery dates, and joined to daily heating degree days (a count of how far each day's average temperature fell below a set base) from the nearest weather station. Then a scored contest between the industry's usual "K-factor" method, a number for each customer worked out from past deliveries that ties gallons used to degree days, and machine-learning forecasts, on the date each tank will fall to 30% full.
The briefCredit · unclaimed02
Winter test problems for routes
A public set of propane delivery-planning problems built from real places: customer points taken from OpenStreetMap buildings in rural counties where many homes heat with propane, tanks of 100 to 1,000 gallons, daily use driven by a real winter's temperatures, bobtail-sized trucks and drive times from a road-network router, with a checker that scores run-outs, miles driven and gallons per stop, and baseline results from open solvers.
The briefCredit · unclaimed
Open 01 · the brief
Score forecasts on real tanks
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of daily tank readings from at least 100 homes that already have remote tank monitors, donated with the owners' consent together with tank size, a coarse location and delivery dates, and joined to daily heating degree days (a count of how far each day's average temperature fell below a set base) from the nearest weather station. Then a scored contest between the industry's usual "K-factor" method, a number for each customer worked out from past deliveries that ties gallons used to degree days, and machine-learning forecasts, on the date each tank will fall to 30% full.
Why it is still open
The readings sit with retailers and monitor makers. Software for delivery companies still sets a K-factor for each customer from the last deliveries and lets staff adjust it by hand, by Oil & Energy's account. The open-source Home Assistant add-on for Otodata monitors reads only a tank's current state and imports no history from the maker's app, and a homeowner app such as PropaneTracker keeps readings typed in by hand on the phone. Searches of GitHub, Crossref and the web in this session found no public set of residential propane tank readings.
Built with what exists
Home Assistant (open-source home-automation software) with the Otodata add-on or an open R3D dial sensor to log readings; NOAA daily temperature records; Python with pandas, scikit-learn or LightGBM.
First step
write a consent form and a Home Assistant export script, and recruit 20 households with monitors through a home-automation forum this month.
Done looks like
at least 100 tanks with one full heating season each, published under an open licence, and a table giving each method's mean error in days to 30% and the share of tanks that would have fallen below 20% before the forecast delivery date.
Open 02 · the brief
Winter test problems for routes
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of propane delivery-planning problems built from real places: customer points taken from OpenStreetMap buildings in rural counties where many homes heat with propane, tanks of 100 to 1,000 gallons, daily use driven by a real winter's temperatures, bobtail-sized trucks and drive times from a road-network router, with a checker that scores run-outs, miles driven and gallons per stop, and baseline results from open solvers.
Why it is still open
The standard public test set for this kind of planning, called inventory routing, is Archetti and colleagues' 2007 set, hosted by Leandro Coelho; it gives each customer the same use in every period, places customers at made-up coordinates and runs for 3 or 6 periods. The closest weather-driven work, Bertsimas, Gupta and Tay (MIT, 2016), made heating-oil use rise as the temperature fell, tuned on Boston weather from November 2013 to March 2014, but put customers at points from TSPLIB, a classic travelling-salesman test library, and gave every customer an identical tank. A 2024 set of 270 "real-world-like" problems in the European Journal of Operational Research varies fleets, horizons, demand patterns and where customers sit, with costs like those found in practice, but its abstract mentions neither weather nor road networks. No released set with real road networks and propane tank sizes was found.
Built with what exists
OpenStreetMap and OSRM (an open road router); US Census tables of house heating fuel to choose counties; NOAA daily temperatures; Google OR-Tools and PyVRP (open routing solvers).
First step
build one county problem from public data and solve its first two weeks with OR-Tools.
Done looks like
at least 10 county problems of 200 to 2,000 customers over a 150-day winter, published with the checker and with baseline scores from at least two solvers that others can rerun.