The Maverick weeding robot from Dutch robot builder Odd.Bot will appear in a GPS-equipped revision at Agrotechniek Holland (ATH), 16-19 September in Biddinghuizen, Netherlands. The update adds a GPS cross-check, a second digital fence, and a rebuilt control app.

What does the new GPS actually do?

The GPS layer acts as a verification tool, not as the primary guidance system. The Maverick still uses its on-board camera to recognise ridges and steer along them, unchanged from the previous build. The GPS signal now independently confirms that headland turns complete properly and that the machine continues in the intended direction along each row.

For a row-crop operation this redundancy matters. A camera-only system can drift when sunlight changes at the headland or when shadows stretch across the row. A satellite reference gives the robot a second opinion on position before any committed turn.

How does the second geofence change field behaviour?

Odd.Bot now lets operators set two separate digital fences in the app:

• Field boundary fence, the standard operating perimeter
• No-turn fence, an inner boundary that blocks rotation inside the crop area

The second fence matters because the Maverick previously mistook a gap in the crop for the field end and turned mid-ridge. The inner boundary lets the robot check whether ridges remain visible before it commits to any rotation.

For irregularly shaped plots, the inner boundary gives growers a way to keep the robot away from obstacles, ditches, or wet patches inside the field without rebuilding the perimeter fence.

What changed in the control app?

Odd.Bot rebuilt the control application around geofence configuration. Every boundary and working-area parameter now sets up directly through the app, replacing the previous mixed workflow that combined software and physical interfaces. The Dutch builder has also folded in several smaller usability tweaks.

The single-app setup cuts training time for new operators. It also reduces the chance of mismatched parameters between the robot, the boundary map, and the user, a common source of faults in earlier weeding robots from multiple vendors.

Can the weeding units operate independently?

Yes. Each weeding module mounted beneath the chassis carries its own camera and control software. Once connected to a 24-volt supply, it works standalone, without the host robot.

Odd.Bot intends to sell these independent modules to third-party machinery manufacturers. Exploratory discussions are underway with several market parties, although nothing concrete has emerged yet. If a deal closes, the 24-volt module could land on inter-row cultivators, hoes, or transplanters built by other makers, giving those brands an autonomous option without developing their own camera stack.

Why does the update matter for cultivators?

Mid-row turning is an expensive fault for vegetable growers running camera-guided inter-row weeding. It damages crop, halts the machine, and demands an operator visit. The combination of a GPS cross-check and a hard no-turn boundary addresses that failure mode without requiring changes to row spacing or the existing weeding mechanics.

For buyers, the 24-volt modules also signal a retrofit path. Existing implements could carry Odd.Bot's camera and actuation technology once a sales channel opens, reducing the upfront cost of going autonomous. A new component-revenue stream for Odd.Bot would also give growers more machine choices across the European weeding market.

What should farmers and integrators watch next?

Watch for the first standalone 24-volt weeding modules shipped to a third-party machinery maker, which would confirm Odd.Bot's pivot from robot builder to component supplier. Monitor how the dual-fence app performs on irregularly shaped fields and short headlands, where the original mid-ridge fault originated. Pricing and timing of first integrations should surface at the next major European machinery shows.