Motion perception for PX4 inspection drones.
We're developing an event-camera perception layer for high-contrast, GPS-denied inspection. Built to work alongside your autopilot, not replace it.
Seeking OEM and integration partners for scoped, paid feasibility work. Implementation is starting. Nothing has flown.
From a dark interior into daylight.
Inspection aircraft encounter abrupt lighting changes close to structures, often without a reliable GPS signal. We are investigating motion perception through those transitions.
Event cameras report changes in brightness rather than complete frames. That makes them an interesting candidate for fast motion and high-contrast scenes, but the advantage has to hold up against your current system.
The starting point is one costly failure, one operating envelope, and a comparison that tells your engineering team whether further integration is worth it.
A defined question. A paid evaluation.
For drone OEMs and robotics integrators with an active inspection program, an existing perception baseline, and a specific problem to investigate.
Scope the decision
Agree the scene, data access, hardware constraints, baseline, and what would justify an integration.
Measure the approach
Develop and evaluate the motion-perception prototype on recorded event data. Include resource use and failure cases.
Make the next call
Receive a reproducible report, an evaluation artifact where licensing permits, and a go/no-go recommendation.
What we agree before starting
A fixed scope, fee, deliverables, acceptance criteria, and schedule. Data availability and engineering capacity are confirmed before a commitment. Hardware and integration are scoped separately.
What this is not
Not an off-the-shelf navigation SDK, a flight-ready system, or a guaranteed performance improvement. A well-supported negative result is a valid outcome of the evaluation.
A testable approach. Not a proven flight system.
We have not published Haltere performance measurements or demonstrated flight. The graphics on this site are illustrations, not product demos.
Our first evidence will be reproducible comparisons on customer-relevant recordings: motion estimates, invalid outputs, processing cost, and the conditions where the approach fails.
Sensor specifications and results from the literature are not Haltere results. Offline replay will not be presented as proof of flight safety or live control-loop performance.
- Useful
- Error against reference measurements
- Robust
- Failures across the agreed operating envelope
- Practical
- Compute, power, hardware, and integration cost
Acceptance thresholds are agreed for each evaluation, not invented for this page.
Development sequence
Updated
- 01
Specification and architecture
CompleteThe detector approach and boundary with the autopilot are specified.
- 02
Core implementation and replay
StartingBuild the Rust motion-perception core and evaluate it on recorded event data.
- 03
Customer-relevant comparison
PlannedCompare against an agreed baseline, including failure cases and total system cost.
- 04
Reference integration
PlannedOne camera and compute configuration. Bench and non-commanding integration before flight.
Sequence, not committed dates. Progress to flight depends on evidence, appropriate facilities, and safety readiness.
Your autopilot stays in control.
The first product scope is motion perception. Recorded event data goes in; motion features, timestamps, and explicit validity indicators come out.
A later reference integration will target one event camera, one companion computer, and one PX4 multirotor configuration. Hardware selection, driver rights, and availability are still to be confirmed.
Flow is not position or a complete navigation solution. The correct interface, coordinate frames, timing, and any additional sensor inputs must be established before feeding estimates to PX4.
Interface validation before integration
Local computation, inspired by motion vision.
The proposed spiking motion detectors correlate neighbouring events without first assembling a full image. The idea draws on insect motion vision; the engineering question is whether it delivers useful perception under a real hardware budget.
Rust supports the intended memory-safety and resource-control requirements. Timing, power, and accuracy still have to be measured.
Architecture, limitations, and references in the technical briefA focused entry point into onboard autonomy.
Haltere is a product of ABX LIMITED, registered in England and Wales. We are building toward reusable perception software for aircraft manufacturers, rather than another airframe or autopilot.
The commercial path starts with paid feasibility work, followed by separately scoped integration. OEM deployment licensing is the intended product model, conditional on repeatable technical value and customer demand.
Company, commercial model, and milestonesEvidence before scale.
The next milestones are a customer-relevant benchmark, a paid follow-on decision, and a reusable reference integration. Those are goals, not traction claims.
Read the investment contextWhat is your aircraft struggling to see?
Tell us about your inspection use case, current platform, and the failure you want to investigate. We can then assess fit and scope a paid evaluation.
Please do not send confidential logs or flight data at this stage. Data access and confidentiality are agreed before an engagement.
What happens after I enquire?
We review the use case, clarify the baseline and constraints, and discuss whether a bounded evaluation makes sense. There is no purchase or delivery commitment until a scope and terms are agreed.
Can I deploy Haltere today?
No. It is pre-release, implementation is starting, and nothing has flown. Evaluation work is not permission for production or autonomous use.
How do fees, licensing, and IP work?
Evaluation fees and rights are agreed in a written scope. Data ownership, confidentiality, and reusable technology are addressed before work starts. Production licensing and ongoing support are separate future agreements.
Is this a replacement for PX4?
No. PX4 retains aircraft authority. The first scope is motion perception; guidance and flight integration require further evidence and safety work.
Evaluation enquiry
This form prepares an email in your email app. You will need to send it there. All four fields are required.
The form needs JavaScript. Please use the direct email link below instead.
Or email hello@haltere.xyz directly.