Skip to content

04 Research & Development

Where the
research happens.

A decade of applied robotics research in autonomous flight, ground autonomy, control and perception. The Lab is not a side project. It is the engine behind every system Altax ships.

Origin & advantage

Most robotics companies buy their autonomy. We wrote ours, in the lab, one control law and one perception pipeline at a time.

Slung-load stabilization, on-board computer vision, ground autonomy, multi-agent coordination, wind-tunnel-validated dynamics. The research came first, and the same autonomy stack now powers our aerial systems, ground robots, and the software behind them. That heritage is the unfair advantage: when a mission needs a manoeuvre no off-the-shelf autopilot can fly, we have already solved it on the bench.

10+ Years of autonomous-flight R&D
4 Active research domains
6 Degrees of freedom, tracked
±2 cm Ground resolution, in service

Research themes

Four fronts,
one autonomy stack.

Each thread is an active line of work, and a proof that the theory leaves the whiteboard and flies.

R-01

Flight Control & Dynamics

The hard part of flight is what happens after the motors spin. We design and validate the controllers that keep a vehicle stable when the payload swings, the wind shifts, or the airframe changes shape mid-mission.

  • Nonlinear control
  • Slung-load
  • Aeroelastic
R-01.01 Swing-free trajectory
R-01.02 Cable-suspended load
R-01.03 Aerodynamic validation
R-01.04 VTOL / fixed-wing

R-02

Computer Vision & Perception

A drone is only as useful as what it understands. On-board detection, motion segmentation and visual tracking turn raw pixels into targets, tracks and measurements, in real time, at altitude, on constrained hardware.

  • Detection
  • Optical flow
  • Fiducials
R-02.01 On-board detection
R-02.02 Motion analysis
R-02.03 Colour tracking
R-02.04 Ground-target follow
R-02.05 ArUco target following
R-02.06 AR fiducials

R-03

Autonomy & Coordination

From a single vehicle holding a centimetre-tight hover to fleets flying a coordinated plan, autonomy is the layer that removes the pilot from the loop, and the ground station that keeps a human commanding the mission.

  • State estimation
  • Multi-agent
  • GCS
R-03.01 Autonomous hover
R-03.02 Indoor navigation
R-03.03 Multi-agent coordination
R-03.04 Swarm formation flight
R-03.05 Mission control

R-04

Optimization & Learning

Some trajectories can’t be hand-tuned. We use trajectory optimization and learning-based control (echo-state networks trained under CMA-ES) to discover aggressive, feasible manoeuvres that a classical controller would never find.

  • Trajectory opt.
  • ESN
  • CMA-ES
R-04.01 Trajectory optimization
R-04.02 Swarm simulation

Flight log

Ten years,
on record.

A mosaic of the missions, benches and simulations behind the research. Every cell is a real flight or test from the archive.

001 HEXA / AIRFRAME
002 FIELD / LOW PASS
003 TUNNEL / 12 m·s⁻¹
004 DETECT / HIGHWAY
005 MESH / FLYTHROUGH
006 TRACK / KINGBEE
007 GCS / TELEMETRY
008 CTRL / SLUNG-LOAD
009 LAUNCH / CATAPULT
010 LAB / INDOOR
011 MOTION / CONTOUR
012 SIM / MULTI-AGENT
013 SWARM / 3-DRONE FORMATION
014 AR / MARKER
015 TRACK / ARUCO FOLLOW
016 AUTO / HOVER
017 OPT / TRAJECTORY
018 TRACK / COLOUR
019 CTRL / SWING-FREE

From the field & bench

Where the theory
meets the field.

Three quadrotors on landing pads in a grass field, staged for a swarm flight
Swarm · launch pads
Three hexacopters lined up on a field table before a coordinated flight
Swarm · pre-flight
Ground-station view of an ArUco marker locked by the tracker with target vector and telemetry
ArUco · lock & track
Long-exposure light painting of a drone's LED flight path traced against the dark

Long exposure · 30 s

Every controller we write eventually draws a line in the air.
This one held its trajectory to the millimetre.

FIG.11 · LED FLIGHT-PATH / AUTONOMOUS TRACKING

Collaborate

Have a problem that
doesn't have an autopilot yet?

We partner with research groups, integrators and operators on applied robotics, from a single hard control problem to a full autonomous system. If it flies, senses or decides, we want to hear about it.