Connectix brAIn: the reader that tells the load from what is around it
Connectix brAIn is an AI filter that runs on the reader itself. It looks at each read in context: what stands still in the field, what moves together with the load, and which way the load went. It sets unwanted reads apart without changing power, antennas or layout. Today you can see it working in the Identix simulators.
The problem: reads that should not count
A read field does not stop at a painted line. At a store exit, a dock door or a conveyor, the reader also reaches the rack beside the door, the pallets waiting in staging or the shelf across the aisle. Those tags were not part of the pass, but they are read as if they were. The usual fixes are physical: less power, an RSSI filter, more distance, shielding. They work, but each one also takes margin away from the load you want to read.
How brAIn filters unwanted reads
The reader does not only see tags: it sees how each tag behaves over time. brAIn uses that behaviour, which a fixed threshold cannot see.
- load
- rack (still)
- piece taken away
- tag at the edge
- 1Before the pass, the reader keeps working: tags that stand in the field all the time, like the rack by the exit, are recognised as residents.
- 2The load passes: its tags rise and fall together. A piece taken from the rack now behaves like the load; the tag on the far shelf shows up only once.
- 3The decision: the load is accepted, the still rack is set apart, and a single isolated read stays uncertain instead of counting as read.
- It learns who lives there. The reader keeps working between passes, so tags that stand in the field all the time are recognised as residents.
- A resident only counts when it changes behaviour. If a piece is taken from the rack and carried out, its signal rises with the load and it counts again.
- One isolated read is not enough. A tag seen only once at the edge of the field is marked uncertain, not read.
- The load is judged as a group. Tags that rise and fall together in time form the load; a tag that does not follow the group is suspect.
Nothing disappears silently: what brAIn sets apart stays marked, so it can be checked.
How brAIn tells the transition between zones
A store exit, a dock door or a corridor joins two zones. Each side has its own antennas, and the reader follows the signal of each tag on each side over time. The side that peaks first is where the item came from; the side that takes over is where it went. An item that only comes close and turns back never lets the other side take over, so no exit is reported.
- side A antennas
- side B antennas
- rack (still)
- 1A person carries an item from the store to the exit: side A reads it first, side B takes over afterwards. The order of the peaks tells the direction.
- 2Another person comes up to the gate and turns back: only side A peaks. The item came close but did not change zone, so no exit is reported.
- 3The rack in zone A is read all the time with no peak: it is resident and never becomes a transition.
- Direction comes from the order of the signals, not from a photocell or any external sensor.
- Tags that move together vote together: a weak tag of the same group follows the direction of the group.
- Residents never become transitions: a rack beside the gate is read all day and never 'leaves'.
Why it is AI and not just one more filter
It learns the place
What stands still in each installation is learned from the reads themselves, and it relearns on its own when the layout changes.
It decides by context
Each read is judged against the history of the tag and against the rest of the load, not against a single fixed number.
It runs at the edge
The decision is made in the reader, read by read, with no cloud and no extra sensors. Power, antennas and layout stay as they are.
What it does not do
brAIn separates tags by how they behave. When an unwanted tag behaves exactly like the load, no algorithm can tell them apart, and the answer stays physical: antenna choice, position, distance or shielding.
- Two dock doors loading at the same time: the pallet at the next door moves just like yours.
- Boxes in a queue on a conveyor and vehicles in a queue or in the next lane: the next one has the same physics as the one being read.
- It does not make an unread tag read. Metal, liquids and dense loads remain a matter of antennas, power and tag placement.
- Gains in the field are expected to be smaller than in simulation: people handling the goods, reflections and tags that appear only now and then reduce the effect.
Connectix brAIn is shown today as a demonstration in the simulators. Every number they show is a simulation, with its premises on screen.
See it in the simulators
Open a simulator, pick a scenario with goods standing nearby (a rack by the exit, a shelf beside the dock, staging next to the tunnel) and turn on the Connectix brAIn switch in the reader settings. The 3D scene shows in grey the tags it set apart.
Questions we hear
Does Connectix brAIn need a photocell or another sensor?
No. It decides only from what the reader already measures for each read. Sensors can be used to check the results, never as an input.
Does it change the reader power or the antennas?
No. Power, antennas and layout stay as they are. brAIn works on the reads, after the radio.
Does it solve every stray read?
No. It solves the cases where the unwanted tags behave differently from the load, such as goods standing nearby. When they move like the load, such as the next dock operating at the same time, the solution is physical.
Can I use it on my readers today?
Today Connectix brAIn is a demonstration in the simulators. If your project has unwanted reads, talk to us: we will look at the layout and at what brAIn could add.