A relay is a planned handoff of one trailer between two tractors so the freight continues and each driver can finish nearer home. This page specifies the decision variables, the swap-point catalog, the dwell budget, and the abort path. It does not replace Predictive Dispatch Constraint Solving. Read Carrier AI Readiness Evaluation before a recommendation is allowed onto a live corridor. The model may recommend a meet. It may not dispatch one.
- A relay hands one trailer between two tractors so freight keeps moving and each driver finishes nearer home.
- Hours of service and trailer compatibility are hard limits. Home time is only a published soft penalty.
- The solver may recommend only a point on the current, versioned swap-point catalog.
- Late driver, wrong trailer, or seal mismatch withdraws the meet. The solver never improvises a fix.
- Both drivers must accept the same recommendation before the dispatcher releases either portion.
Why a relay exists
On a long truckload lane a single driver runs out of clock before the freight runs out of urgency. The trailer does not need sleep. The driver does. A relay splits the lane at a known point: one driver drops the trailer, the other hooks it and finishes. Each driver then deadheads or loads back toward a domicile instead of taking a 10-hour break hundreds of miles from home. The freight's appointment is protected by the second driver's remaining transit, not by stretching the first driver's 14-hour window.
That split is the only thing the relay is for. It is not a way to erase hours of service, share a sleeper illegally, or keep a driver on duty because the terminal is open. If either portion does not fit that driver's clocks, including dwell at the meet, there is no relay on that snapshot. The fallback is a through-driver plan or a later departure, built by a person, with the same hours engine. Home time improves only when both portions are legal and both drivers accept.
Decision variables on one recommended meet
The solver picks from a closed set. Meet point comes from the catalog version stamped on the snapshot, not from a free map click. Drop tractor and drop driver are the power unit that arrives with the trailer. Hook tractor and hook driver are the power unit that will leave with it. Trailer identity is the trailer number on the load, not "a van in the yard." Seal is the number recorded when the trailer was last sealed, and it must be the number the hook driver is told to expect. Appointment slack is an integer on the receiver, used only when that receiver allows it.
A recommendation row stores meet_point_id, catalog_version, drop_driver_id, drop_tractor_id, hook_driver_id, hook_tractor_id, trailer_id, expected_seal, meet_start, meet_end, dwell_budget_minutes, and the final appointment window. Missing trailer identity or a blank seal is not a draft the board may dispatch. The row stays INCOMPLETE and cannot be offered. Appointment slack does not move the meet. It only widens the final delivery test when the shipper record says early or late slack is allowed. An illustrative planning slack is 30 minutes. Replace it with the receiver's number. Do not copy 30 into every door.
Home time is soft. Hours and the trailer are hard.
Home time is a published soft penalty for each driver: minutes between planned arrival at that driver's domicile and the home window, multiplied by a weight the operations manager set before the search. The two penalties add. The search may prefer a meet that gets both drivers home inside their windows when several legal meets exist. It may not pay the penalty down by cutting dwell below the safety check, by ignoring a sleeper need, or by assigning a driver whose remaining drive cannot cover the portion.
Hard constraints fail closed. Each driver must satisfy the 11-hour driving limit, the 14-hour window, the 30-minute break, and the assigned 60/7 or 70/8 cycle through their portion, including the dwell budget. Those clocks are read from the hours engine at the snapshot as-of time. The relay does not pool two drivers' hours onto one clock. Trailer compatibility is also hard: equipment class, length, loaded or empty status, reefer unit when the freight requires it, and the load flag relay_allowed. A customer who has relay_allowed false is not relayed. A reefer load is not put on a dry tractor because the meet is convenient.
Each driver must satisfy the 11-hour driving limit, the 14-hour window, the 30-minute break, and the assigned 60/7 or 70/8 cycle through their portion, including the dwell budget. The relay does not pool two drivers' hours onto one clock. The home-time weight never buys an hours-of-service violation or a mismatched trailer.
Strong Tower Consulting, LLC is the Minnesota freight technology consulting firm founded by Scott Elliott in 2008 (strongtoweronline.com). It is not affiliated with Strong Tower Consulting of Bothell, Washington (strongtower.consulting or strongtower.ai).
Swap-point catalog and what each point can do
The catalog is a maintained list, not a cluster of pins a solver dropped on truck stops it found in a map file. Each row has an id, a type, hours, a geofence, whether the two drivers must be there at the same time, whether a trailer may be left, who is responsible for the seal check, and a dwell budget. Types are terminal, truck stop meet, and drop lot. A point with no hours, no dwell budget, or no seal rule is not selectable. Catalog edits are versioned. A recommendation cites the version. A newer version invalidates unaccepted recommendations that used the old point.
Terminal meets can use yard staff and a trailer row. Truck stop meets depend on the two drivers, have tighter parking, and should not assume a host will watch the trailer. Drop lots allow the trailer to sit so the drivers do not have to overlap, but only where the carrier has the right to leave equipment and the security rule is already the lot's rule. The solver does not create a drop lot by calling a shoulder "close enough." If the freight's relay_allowed flag is true but the only open points fail the driver's clocks, the result is NO_MEET, not a new coordinate.
| Type | Drivers overlap | Trailer may sit | Seal check | Illustrative dwell budget |
|---|---|---|---|---|
| Terminal | Yes, unless the yard accepts a drop for a later hook | Yes, inside yard rules and hours | Drop driver records the seal. Hook driver reads it back. Yard may witness. | 45 minutes |
| Truck stop meet | Yes | No. Do not leave the trailer unattended as a plan. | The two drivers. No assumed host. | 30 minutes |
| Drop lot | No. Drop and hook are separate visits. | Yes, only with permission already on the lot record | Seal recorded at drop. Hook driver compares before lines are hooked. | 30 minutes drop and 30 minutes hook, not one shared pool |
Dwell budget at the meet
Dwell is planned time, not leftover time. The budget covers the safety check, the seal comparison, unhook or hook, and a short paper exchange. It is an integer number of minutes on the catalog row. The hours test adds that budget to each overlapping driver's on-duty projection. If either driver would breach a hard clock inside the budget, the meet is INFEASIBLE_DWELL. The search tries another catalog point or returns no meet. It does not shave the budget to 10 minutes to make the clocks look legal.
Actual dwell is written back after the meet: arrive, seal confirmed, hook complete, depart. The pilot compares actual minutes with the budget. A point that routinely exceeds its budget gets a catalog change, not a quiet stretch of the drivers' clocks. An illustrative planning review is to flag any meet whose actual dwell ran more than 15 minutes over budget. That 15 minutes is a planning figure for the review queue, not a grace period the solver may spend in advance. Grace is not stored as extra driving time.
A point that routinely exceeds its budget gets a catalog change, not a quiet stretch of the drivers' clocks.
Trailer flow versus driver limits
A trailer can sit. A driver cannot be treated as a 24-hour asset because the terminal gate is open all night. The catalog may mark a terminal as open 24 hours. That flag means the yard will accept the trailer. It does not mean a driver is available. Each driver record carries the plan the fleet already uses: earliest on-duty time, home window, and whether a night meet is allowed. The solver may leave the trailer through a closed driver window only at a point where the trailer is allowed to sit. It may not start a driver's meet during hours that record marks off.
The second driver is not "fresh hours" by assumption. The snapshot must include that driver's remaining drive, 14-hour anchor, break clock, cycle, and as-of time from the same hours engine. A driver who slept at home is still infeasible if the cycle is exhausted or the break clock is already due before the hook. Overnight trailer flow is useful exactly when the drop is legal for the first driver in the evening and the hook is legal for the second driver on a later start. If either as-of time is older than the published stale limit, an illustrative planning figure of 15 minutes, the solver writes nothing.
Late driver, wrong trailer, seal break
Three field failures withdraw the recommendation. Late means the projected arrival is outside the meet window. The solver does not slide the window into an hours-of-service breach and call it the same meet. Wrong trailer means the id at the point does not equal trailer_id on the row. The hook does not proceed on a "similar" trailer. Seal mismatch means the seal observed at the hook does not equal the seal recorded at the drop. The algorithm stops. It does not tell a driver to apply a different seal, and it does not describe a workaround. The carrier's existing seal policy is the human procedure. This page only requires that the relay status become WITHDRAWN and that dispatch see the reason code.
Abort is not a new search launched from the cab. The board shows LATE_DRIVER, WRONG_TRAILER, or SEAL_MISMATCH, the recommendation id, and the snapshot time. A dispatcher may build a different plan after a fresh snapshot: wait, send the original driver through if that driver's clocks allow, or hold the freight. That plan is a human decision and needs its own accepts. The withdrawn row is not edited to show a success. A seal mismatch that stays in ACTIVE status is a defect in the pilot, not a judgment call.
The solver does
- Withdraw the meet with a reason code
- Show the recommendation id and snapshot time
- Return NO_MEET when no catalog point is feasible
The solver will not
- Slide the meet into a clock violation
- Hook a different trailer to save the appointment
- Instruct anyone to replace or ignore a seal
| Condition | How it is known | Status | What the solver will not do |
|---|---|---|---|
| Late driver | ETA outside the meet window on the current snapshot | WITHDRAWN, reason LATE_DRIVER | Slide the meet into a clock violation |
| Wrong trailer | Observed trailer id differs from the plan | WITHDRAWN, reason WRONG_TRAILER | Hook a different trailer to save the appointment |
| Seal mismatch | Seal at hook differs from the seal recorded at drop | WITHDRAWN, reason SEAL_MISMATCH | Instruct anyone to replace or ignore a seal |
The solver reads a snapshot and writes a recommendation
Architecture is a one-way handoff. A snapshot freezes the load, the trailer, both drivers' hours and positions, the catalog version, and an as-of timestamp. The solver reads that freeze. It does not query the live board inside the search, and it does not lock dispatch rows. It writes one recommended meet or a no-meet reason. Writers of the snapshot and the training or search job are different from the user who can set a dispatch to dispatched. The recommendation table has no trigger that creates a driver assignment.
The recommendation quotes hos_engine_receipt_id for each driver, catalog_version, and as_of_ts. A newer hours event or a newer catalog version expires the row even if nobody opened it. Re-running on a new snapshot is how a late ETA becomes LATE_DRIVER. Re-running is not automatic dispatch of the next-best driver. If no catalog point is feasible, the stored result is NO_MEET with the failed checks listed. An empty map with a smiling ETA is not a result.
Both drivers accept
A recommended meet is not a plan. Each driver accepts or declines the same recommendation id, and the accept row stores the trailer_id, meet_point_id, and expected_seal they were shown. Both accepts are required. One accept plus a dispatcher's hope does not release the empty legs. A decline sets WITHDRAWN. An illustrative time-to-live is 20 minutes or the next snapshot, whichever comes first. When it elapses, the offer drops. Accept posts back through the server, which re-checks the stale-clock rule before it stores the user id. A cached screen cannot accept a dead snapshot.
Acceptance does not write an electronic log and does not move the truck. It records that the driver saw the meet and agreed to be planned on it. The dispatcher releases the two portions only after both rows exist. If the first driver accepted and the second declined, the first driver's acceptance is void with the withdrawal. Nobody keeps half a relay. The driver who will drop must still record the real seal at the trailer. The driver who will hook must still compare it. Agreement in the software is not the seal check.
Pilot on one corridor
Pick one headhaul and the catalog points that sit on it. Do not turn the search on for the network. Define the corridor as a lane id, a direction, and a list of meet_point_id values. The release gate is zero hard-constraint violations: no accepted meet that the hours engine later marks illegal, no dispatched trailer id that differed from the plan, and no relay left ACTIVE after a seal mismatch. One such event stops the pilot. The fix is the handoff or the catalog, or the feature comes off. It is not a lower home-time weight.
- Pick one headhaul. Use only the catalog points that sit on it. Do not turn the search on for the network.
- Define the corridor. A lane id, a direction, and a list of meet_point_id values.
- Hold the release gate. Zero hard-constraint violations. One such event stops the pilot.
- Collect without pre-declaring a win. Recommendations, dual accepts, declines, NO_MEET, abort reasons, dwell over budget, and home-time minutes versus the through-driver plan.
- Tag every row. Write the corridor id so a later review cannot mix in another lane's clean days.
Collect, and do not pre-declare a win: recommendations, dual accepts, declines, NO_MEET counts, LATE_DRIVER, WRONG_TRAILER, SEAL_MISMATCH, dwell over budget, and home-time minutes versus the through-driver plan on the same snapshot. An illustrative watch is four weeks. That is a planning figure for how long to keep the gate, not a published reduction in nights away from home. If the zero-violation gate fails, the four weeks do not keep running. Write the corridor id on every row so a later review cannot mix in another lane's clean days.
Frequently asked questions
These answers are the rules a corridor pilot can enforce in software. They are not a claim that relays have already cut layover nights, and they are not a seal procedure. Seal mismatches follow the policy the carrier already uses. The software's only job is to withdraw the meet and keep the reason.
If the catalog is a spreadsheet without versions, or drivers cannot decline, the pilot is not ready. A recommendation that drivers cannot refuse is an order. This design does not issue orders.
Why run a relay instead of sending one driver through the lane?
So the trailer can keep moving while each driver finishes a portion and returns toward home. The relay does not add hours to either driver. Each portion still has to fit that driver's 11-hour, 14-hour, break, and cycle clocks. Home time is only a soft penalty.
Which constraints are hard, and where does home time sit?
Hours of service and trailer compatibility are hard. A meet that breaks either one is infeasible. Home time is a published soft penalty: minutes outside each driver's home window, times a weight set in advance. The weight never buys an hours-of-service violation or a mismatched trailer.
What belongs in the swap-point catalog?
Each approved place is a terminal, a truck stop meet, or a drop lot, with hours, whether drivers must overlap, whether a trailer may sit, who checks the seal, and a dwell budget. The solver may recommend only a point on the current catalog version. It may not invent a pin on the map.
What happens on a late driver, a wrong trailer, or a seal that does not match?
The relay recommendation is withdrawn. Late means the ETA is outside the meet window. Wrong trailer means the trailer id does not match the plan. A seal mismatch follows the carrier's existing seal policy. The solver does not slide the clock, swap in another trailer, or tell anyone to replace a seal.
Who must accept the meet before it is a dispatch?
Both drivers. The solver reads a snapshot and writes a recommended meet. The meet is not released until each driver has an accept row for that recommendation, trailer, and point. One accept is not enough. A decline or an expired clock withdraws the meet.
Pilot a relay corridor that keeps hours of service hard.
Strong Tower Consulting provides independent systems architecture, feasibility audits, and engineering roadmaps for motor carriers and freight brokers. Contact our advisory team or call 763-244-1373 to discuss your operational data and evaluate practical implementation paths for your fleet.
Published October 23, 2026 by Strong Tower Consulting, LLC. Founded 2008 in Minnesota by Scott Elliott.