Intelligent Task Dispatch Engine

Polaris

autonomous factory orchestration.

Polaris is the reinforcement learning engine behind every dispatch decision. It builds an execution strategy from your operational history and from training in simulation on that same data. It helps operators make the best long-term decisions across multiple goals, and can run the operation autonomously.

How it works Contact ↗
0
Tasks performed
23%
Utilization increase
15%
Reduction in temperature loss
0
Hours of live operations
At a glance
built for complex decision making

Executable options, not just the optimal one.

Only 29% of the recommended actions are followed with classical planning tools. Polaris closes that execution gap: 71% of its recommendations are acted on in live production. It learns from past experience and trains in simulation on your own historic data, so its plans are both executable and strategic. It supports your operators, and where you want it to, it runs the operation autonomously.

29%classical planning, baseline follow rate
71%Polaris, measured in live production
Replanning
<1 s
Things change on the floor all the time, and nobody can wait for a new plan. Polaris re-evaluates the whole plan and answers within a second.
Strategy
Multi-goal
Strategic plans that weigh several goals at once across the whole shift, not just the best next move.
Autonomy
Your call
Polaris advises your operators, or runs the operation autonomously through Compass. The same engine, at the level of autonomy you choose.
Capabilities

Learn. Plan. Execute.

Three layers that work on every decision together: what Polaris has learned from your operation, the plan it builds for the whole shift, and the control your operators keep over it. When something on the floor changes, all three respond at once.

Polaris has studied how your operation actually works: what gets done, what gets skipped, and what it takes to reach plans that hold up on the floor.

→
Learns from operational historyYour own records teach Polaris what your team will actually follow, without a single interview or hand-coded rule.
→
Trains in simulation on historic dataThe simulation is built from your own operational history, not a simplified test environment, so what Polaris learns there holds up on the floor.
→
Rules out the impossible firstPhysically impossible and procedurally forbidden moves are excluded before the search begins. No wasted computation.
→
Plans inside realityRecommendations stay within the range of moves your operators will actually choose.

Polaris is the engine underneath Compass: every autonomous decision, order priority and assignment runs through it. On your floor it can advise your operators, or run the operation itself.

→
Decision supportRecommendations arrive as executable options with the reasoning behind them, ready for the operator to act on.
→
Autonomous decision makingWhere you allow it, Polaris commits the plan itself and the fleet executes through Compass, with no one in the loop.
→
ControlOperators stay in the loop and can override any call on the spot. Every override is a decision the system learns from.
→
Real-timeA delay, a breakdown or a new order, and the next plan is already there. Nobody waits for the schedule to catch up.

Nothing hidden, nothing forced. Every call can be traced and compared against your baseline.

→
Baseline comparisonEvery recommendation is compared with what current practice would have done under the same conditions. If Polaris can't beat your baseline, it says nothing.
→
Traceable recommendationsAny call can be shown as the conditions that produced it, in language the floor can read. Consistent and explainable shift after shift, so operators can lean on it instead of second-guessing it.
→
Strategy grows with trustPolaris starts close to current practice, planning inside what your team already accepts, and extends its strategic horizon as follow-through grows.
How it works

Every order. Decided.

Polaris evaluates the state of the floor continuously. Every incoming order, every delay and every completed move triggers a re-evaluation of the full plan: thousands of scenarios are scored against your baseline and the best executable option is dispatched. The panel below runs that loop.

Polaris · Dispatch Engine On plan LIVE  --:--:--
OrderLocationVehiclePrioStatusETA
71.2%
Follow rate
+18.3%
Optimization vs baseline
11
Queued
98%
On schedule
What it does

Every dispatch decision,
learned from your operation.

Polaris learns how your operation actually runs from your own operational history, explores what could happen next in simulation, and turns both into recommendations your operators can act on.

01
Learned decisions

A model trained on your operational history: what was moved, when, by whom, and what the floor accepted. The judgment your operators built over years becomes part of every recommendation.

02
Explored scenarios

Thousands of scenarios per second are simulated forward before a recommendation is made. The plan that holds up across them wins, not the one that only looks best right now.

03
A strategy, not a next step

A short-sighted planner takes the best move available right now. Polaris weighs the whole shift and accepts a slower move now when that keeps the operation ahead later, across every goal at once.

04
Replanning on change

A delay, a breakdown or a new order triggers a full re-evaluation. Polaris does not fall back to a contingency table; it builds a new plan from first principles in under a second.

Where this goes
”

Changing factories into self-learning factories.

The next decade of industrial AI will be won by the systems that get used, not the ones that are theoretically best.Quantillion
  1. Start with a system operators can follow, built on their own historic decisions.
  2. Find new optima within the physical possibilities of your industry.
  3. Expand across processes into a complete orchestration that adapts to every change.
In use

One engine.
Different floors.

Polaris runs the same reasoning loop wherever tasks compete for time, people and machines. Two operations it is built around:

Example 01 · Steel

Ladle logistics against the clock

A hot-metal ladle loses heat every minute it stands idle. Polaris keeps a live model of every ladle's temperature, every vehicle's position and every furnace's readiness, and re-solves the full routing plan every few seconds. At one steelworks this cut temperature loss by 15%.

  • Thousands of forward simulations before each commit
  • Full replan in under a second after a breakdown
  • Charge timing held for basic oxygen furnaces
  • Minimized heat loss, maximized furnace utilization
Example 02 · Aluminium

One plan across the whole smelter

A smelter runs several fleets at once: crucible carriers, tapping vehicles and transport between potroom, cast house and remelt. Polaris plans them as one operation, holds every temperature window, and picks the sequence that loses the least heat.

  • Several vehicle types planned as one fleet
  • Transport held inside its temperature window
  • Potroom, cast house and remelt kept in step
  • Replans in seconds when a tap runs long

Same learning loop, same follow-rate measurement, same override rules; the process underneath is what changes.

04
Integrated steelworks operations

Steel

Hot-metal ladle logistics, heat-loss minimization, and furnace scheduling in integrated steelworks. Execution order planned to reduce temperature loss and maximize yield.

Polaris · Compass
02
Greenhouse operations

Agriculture

Greenhouse farming scheduled as one operation. Harvesting, packing and transport planned across crews and vehicles, and replanned as yields and orders change through the day.

Polaris
01
Aluminium smelter operations

Aluminium

Orchestration of several fleet types across potroom, cast house and remelt. Heat loss minimized on every transport, and the optimal sequence found for recycled metal.

Polaris · Compass
Get in touch

Let's get your plans
followed.

Contact us ↗ Back to overview →