Skip to content
Hybot

Article

How robot task assignment algorithms work

A task assignment algorithm decides which robot takes which job. Simple rules pick the nearest free unit or serve jobs in arrival order. A weighted approach scores every waiting job against every available robot using several factors at once, and re-runs continuously so the decision reflects the fleet's current state rather than a stale snapshot.

Written by Hybot technical lead, Technical lead, Hyrcan-Tech · · 6 min read

Three ways to decide, and what each costs

RuleHow it decidesWhat it gets wrong
First-comeOldest waiting job to the first free robotIgnores where the robot is; produces long empty trips
Nearest-robotJob to the closest available unitIgnores charge and fairness; exhausts one machine
Weighted scoringSeveral factors combined into one scoreMore parameters to tune, and you have to choose them honestly

The first two are appealing because they are explainable in a sentence. They are also the two that produce the fleet behaviour venues complain about.

Why nearest-robot fails in practice

Consider two robots. One is four metres from the pass at 15% battery and about to hit its charging threshold. The other is twelve metres away at 90%.

Nearest-robot sends the first. It accepts the job, starts the run, drops below its threshold, and now there is a recovery problem where there was a delivery. The eight metres saved cost several minutes and an interrupted task.

Repeat that thirty times in a service and you have the classic symptom: one unit flat by eight o'clock, the other barely used.

Weighted scoring

Hybot scores each waiting task against each available robot using three inputs weighed together:

The important word is together. Applying them as ordered tie-breakers — priority first, then distance, then battery — collapses to a rule where the later factors almost never matter, because exact ties are rare. Combining them into one score is what lets a much better charged robot beat a slightly closer one.

Continuous re-evaluation

Hybot re-scores the entire queue every two seconds. This matters because every input is volatile: robots finish, batteries drop, priorities arrive.

A useful side-effect is elasticity. Adding a robot mid-shift needs no rebalancing step — the next cycle simply has an extra candidate — and robots can be added without restarting the system.

Where the intelligence is, and is not

This is deterministic scheduling, not machine learning, and we describe it that way deliberately. An operator needs to be able to predict which robot takes a job; a learned policy that is right more often but occasionally surprising is a poor trade on a restaurant floor. Our AI cluster sets out where learning genuinely applies in this field.

Where to go next

Cluster hub: Intelligent robotics.

Frequently asked questions

What is wrong with assigning to the nearest robot?

It ignores everything except distance. The nearest unit may be at 15% battery, or may be the one that has done every job this hour. Nearest-robot rules reliably produce one exhausted machine and one that barely moved, which is a worse outcome than a slightly longer trip.

Why re-evaluate instead of assigning once?

Because the inputs change while the job waits. A robot finishes early, another drops below its charging threshold, a more urgent task arrives. Hybot re-scores the whole queue every two seconds, so an assignment reflects the fleet as it is rather than as it was.

How are the factors combined?

Weighed together rather than applied as ordered tie-breakers. Priority, distance to pickup and remaining battery each contribute to a single score, so a marginally closer robot does not automatically beat a much better charged one the way a strict ordering would force.

Where this fits

This page is part of Intelligent robotics: coordinating a fleet. If you are working through the topic in order, these are the neighbouring pages.

  • Multi-robot coordination, explained

    Two robots are more than twice the problem of one. The failure modes that appear at the second unit, and what a coordination layer has to do about them.

  • Robot fleet monitoring and heartbeats

    How a deployed site reports its own health — heartbeats, latency, service status and node redundancy — so problems are found before a customer reports them.

  • Service robot battery life and charging

    Why battery is an input to the dispatch decision rather than an alarm, how automatic charging works, and what charger placement does to a venue's effective capacity.

Take it further

If a question here applies to a venue you actually run, the specifics matter more than the general case.