How to use this document
Each project is a self-contained engineering exercise with a falsifiable completion criterion. A project is finished when the criterion is met and documented — not when it works once on a bench.
The ladder is ordered by dependency, not difficulty. Some projects can run in parallel; the phase gates cannot be skipped. Where a project has a stated failure mode, that failure mode is the reason the project exists.
Three principles govern the whole programme:
- Build the mock structure early A representative tower section at ground level — lattice members at correct spacing and section, a real antenna mount, real fasteners — is the single highest-leverage piece of equipment in the programme. Almost every project from Phase 2 onward is faster and safer with one. Build it before Project 5.
- Every project logs evidence From Project 1, every run records its inputs, outputs and measurements in a consistent format. The commercial product is verified work, and the logging discipline that produces it cannot be retrofitted at Project 19.
- Nothing modifies the structure No rails, no fitted tracks, no pre-installed fixtures. Any solution requiring the tower to be prepared first reintroduces the climb and destroys the value proposition.
Cross-cutting requirements
These apply to every project and are not projects in themselves. They should appear as constraints in each design review.
-
Weight and transport budget
Two people must be able to carry, rig and recover the machine. Set a target mass early and hold every subsystem against it. A machine needing a crane has replaced one logistics problem with a worse one.
-
Dropped object prevention
Every tool, fastener and removable component is tethered or captive. A dropped 10 mm spanner from 40 m is a fatality. This is not a Phase 6 concern.
-
RF exposure and EMC
Work occurs near live transmitting antennas. Electronics must tolerate the RF environment without malfunction, and any operating procedure must respect exposure limits and any required downtilt or power reduction. Characterise this before working on live sites.
-
Weather envelope
Wind, rain, temperature and ice are the field conditions. State the envelope in which the machine is rated to operate, and design the fault behaviour for conditions outside it.
-
Structure variation
Lattice, monopole and guyed masts have fundamentally different climbing geometry. Decide early which is the primary target — lattice is the most tractable and the most common in the markets where development access is likely — and treat the others as later generalisation.
Phase 1
Actuation and safety foundations
Projects 1–4
Gateno climbing work begins until Project 4 is complete.
Project 1. Motor control and drive characterisation
Objective
Drive a geared or brushless actuator under closed-loop current control and characterise its real torque, speed and thermal envelope.
Why it matters
Tower work is intermittent high-torque, which is a thermal problem rather than a peak-torque problem. Everything downstream depends on knowing what the actuator can actually sustain.
Done when
Measured torque-speed curve produced from your own bench data, not the datasheet. Continuous duty cycle at which the motor reaches thermal limit documented. Current limiting demonstrated under stall.
Failure mode
Sizing from datasheet peak torque, then discovering the continuous limit at height with a bolt half-turned.
Project 2. Closed-loop position control under load
Objective
Hold and track commanded position within a stated tolerance while an external disturbance is applied.
Why it matters
The tool must stay engaged with a fastener while the platform moves in wind. Position control that only works unloaded is not position control.
Done when
Holds position within stated tolerance under a step disturbance of at least 50% of rated load. Settles without sustained oscillation. Tuning method and final gains documented and reproducible.
Failure mode
Tuning on a rigid bench with no load and no compliance, then finding the system unstable once mounted on anything that flexes.
Project 3. Force and torque sensing
Objective
Measure applied force and torque with known accuracy, repeatability and drift.
Why it matters
Torque applied to specification is the deliverable and the evidence. Unmeasured torque is commercially worthless.
Done when
Calibrated against a certified reference. Error budget stated. Repeatability documented over at least fifty cycles. Temperature drift characterised across the intended operating range.
Failure mode
Using motor current as a proxy for output torque. Gearbox friction and efficiency vary with load, temperature and wear, and the resulting error is large, non-linear and unacceptable as evidence.
Project 4. Fail-safe grip
Objective
A gripping mechanism that maintains hold on a structural member when power is removed, and that requires a positive command to release.
Why it matters
This is the most important mechanism in the machine. Every other capability is negotiable; this one determines whether the safety case is possible at all.
Done when
Holds full machine mass plus a documented dynamic factor with power disconnected, demonstrated repeatedly. Release requires positive command and cannot occur through single-point failure. Mechanism principle documented — over-centre linkage, spring-applied power-released brake, irreversible worm drive, or equivalent.
Failure mode
Any grip depending on continuous power or continuous control. If a software fault or a flat battery can drop the machine, no operator will permit it on their structure and no insurer will price it.
Phase 2
Manipulation
Projects 5–8
GateBuild the mock structure before starting this phase.
Project 5. Force-controlled gripper
Objective
Grasp and hold objects of varying geometry with controlled, measured grip force.
Why it matters
Components handled on a tower — light units, brackets, radio units — are of differing size and fragility, and grip force must be sufficient without damage.
Done when
Grasps a defined set of test objects with grip force controlled to a stated tolerance. Detects and reports slip. Retains hold on power loss.
Project 6. Multi-axis manipulator
Objective
A manipulator with sufficient degrees of freedom to reach fastener and mount positions from a fixed platform attachment point.
Why it matters
The reachable workspace determines how many times the platform must reposition, and repositioning is the slowest and highest-risk part of any operation.
Done when
Reachable workspace mapped and documented against the mock structure's fastener positions. Percentage of target positions reachable from a single platform attachment stated. Singularities identified and handled.
Failure mode
Optimising for reach at the cost of stiffness. A long compliant arm cannot deliver accurate torque.
Project 7. Calibrated torque tool
Objective
Apply a commanded torque to a fastener to a stated accuracy, and record the value achieved.
Why it matters
This is the core deliverable of the whole machine. Antenna mount fasteners have specified torques and the evidence of compliance is what the operator is buying.
Done when
Applies commanded torque within a stated percentage error, verified against a reference transducer, across the full intended range and across at least fifty repetitions. Records applied torque, angle and time-series for every operation.
Project 8. Reaction force and base stability
Objective
Characterise how manipulator forces transfer into the platform, and quantify the torque at which platform position exceeds tolerance.
Why it matters
Applying 60 Nm to a bolt applies 60 Nm to the platform. Achievable torque accuracy is set by the grip and structural stiffness, not by the arm. This coupling is the hardest problem in the design and the most common integration failure in mobile manipulation.
Done when
Measured relationship between applied torque and base displacement documented. Maximum applicable torque within positional tolerance stated. A reaction compensation strategy — bracing, counter-torque, second attachment point — demonstrated to raise that limit.
Failure mode
Developing arm and platform as independent subsystems and discovering the coupling at integration, by which point both architectures are fixed.
Phase 3
Perception and measurement
Projects 9–11
Project 9. Structural member detection and pose estimation
Objective
Detect tower structural members and estimate their position, orientation and section relative to the machine.
Why it matters
The climbing system must know where the next attachment point is, at what angle, and whether it is within grip capability.
Done when
Detects and estimates pose of members on the mock structure and on photographs of real towers, within stated positional and angular error, under varying light. Reports confidence and fails safe when confidence is low.
Project 10. Antenna and mount detection
Objective
Detect antennas, identify mount type and locate the adjustment hardware.
Why it matters
Mounts vary widely across manufacturers, vintages and undocumented site modifications. The machine must recognise what it is looking at before it can act on it.
Done when
Classifies a documented set of common mount types from real site imagery. Locates fastener positions within stated error. Explicitly reports unrecognised configurations rather than guessing.
Project 11. Azimuth and downtilt measurement
Objective
Measure antenna azimuth and mechanical downtilt with a stated and defensible error budget.
Why it matters
These are the measurements the operator is paying for, before and after intervention. The number is only worth what its stated accuracy is worth.
Done when
Measurements agree with a reference method within a documented tolerance. Full error budget published, including sensor error, mounting error and structural movement. Required accuracy confirmed with at least two tower operators, not assumed — this determines the entire sensing architecture and it is cheap to ask.
Failure mode
Achieving impressive precision against a specification nobody asked for, or discovering late that the industry requires tighter tolerance than the sensor suite can deliver.
Phase 4
Locomotion
Projects 12–14
GateProject 4 must be complete. All work in this phase begins at low height over matting.
Project 12. Controlled ascent with secondary retention
Objective
Ascend a structural member under control, maintaining at least two independent points of attachment at all times.
Why it matters
Two-point attachment is the rule human climbers work to, and for the same reason. A machine that ever has a single point of failure between itself and the ground will not pass a safety case.
Done when
Ascends the mock structure at a documented rate with two independent attachments maintained continuously, verified by instrumentation rather than observation. Demonstrated across repeated cycles with no single-point-of-failure state.
Project 13. Controlled descent and load holding
Objective
Descend under control, and hold position indefinitely at any point without power.
Why it matters
Descent is more dangerous than ascent, because gravity assists the failure mode. Most climbing mechanisms are easier to design going up.
Done when
Descends at a controlled rate with no uncommanded acceleration under any single fault injected. Holds indefinitely at arbitrary positions with power removed. Emergency stop leaves the machine in a stable held state.
Project 14. Obstacle traversal and structure variation
Objective
Traverse joints, bolted plates, cable trays, section changes and other real-world discontinuities. Handle variation in member size and geometry.
Why it matters
Real towers are not clean prisms. They carry decades of undocumented additions, and every one of them is between the machine and the work.
Done when
Traverses a documented catalogue of obstacle types on the mock structure. Member size range handled is stated. Behaviour on encountering an untraversable obstacle is defined, safe and reported to the operator.
Failure mode
Building a mock structure that is cleaner than reality. Photograph real towers and reproduce their mess deliberately.
Phase 5
Remote operation and fault behaviour
Projects 15–17
GateThis phase is what makes the machine deployable. It is also what an operator's risk function will examine most closely.
Project 15. Teleoperation with latency and feedback
Objective
Operate the full machine remotely with visual and force feedback sufficient to perform manipulation tasks.
Why it matters
Teleoperation is the product. Autonomy is a later efficiency gain layered on top of a working service.
Done when
An operator who did not build the machine completes a manipulation task on the mock structure using only the remote interface. Performance characterised across a range of induced latency and bandwidth constraints representative of field connectivity. Camera selection and arbitration usable without training in the mechanism.
Failure mode
An interface only its author can drive. Test with someone else, early.
Project 16. Loss-of-link and fault-state behaviour
Objective
Define, implement and prove deterministic behaviour under communications loss, power fault, sensor failure and command timeout.
Why it matters
This is the first question any competent safety assessor asks. There must be a documented answer with evidence behind it, not an intention.
Done when
Behaviour on each fault class is specified in writing, implemented, and demonstrated by injecting each fault. The machine reaches a defined safe state — attached, held, not moving — within a stated time in every case. Recovery from safe state on link restoration is controlled and requires operator confirmation.
Project 17. Recovery of a disabled machine from height
Objective
A proven method of recovering the machine from the structure when it is unpowered or unresponsive.
Why it matters
Every operator will ask what happens when it dies at thirty metres. "It hasn't yet" is not an answer, and a machine that requires a rigger to climb up and free it has reintroduced the exact risk the product exists to remove.
Done when
At least one recovery method demonstrated on a fully unpowered machine at height — controlled descent on reserve power, mechanical release from ground via tether, or equivalent. Recovery procedure documented to the standard a method statement requires. Time to recover measured.
Failure mode
Treating this as an afterthought. It constrains the mechanism, and discovering the constraint after the mechanism is fixed means redesigning it.
Phase 6
Field tasks and evidence
Projects 18–20
Project 18. Fastener engagement under field conditions
Objective
Locate, engage and turn fasteners that are corroded, painted, seized, non-standard, or out of nominal position.
Why it matters
This is where laboratory manipulators fail. A clean hex head on a bench is a solved problem; a twenty-year-old galvanised bolt with paint over the flats and a mount that has shifted three millimetres is not.
Done when
Engages and turns fasteners from a deliberately degraded test set — rusted, painted, damaged, mispositioned — with a documented success rate. Detects failure to engage and reports rather than forcing. Breakaway torque limit defined so the machine does not shear a fastener it cannot see.
Failure mode
Testing only on new hardware. Buy or make a degraded fastener set and use it from the beginning of this phase.
Project 19. Complete work tasks with evidence capture
Objective
Perform full maintenance operations end to end — azimuth adjustment, downtilt adjustment, bracket manipulation, component replacement — each producing an evidence package.
Why it matters
The commercial product is not the motion; it is the verified completed work order. The evidence package is what the operator files and what their client accepts.
Done when
Each target task completes on the mock structure. Each produces an evidence package containing before and after measurements with stated accuracy, torque values applied against specification, timestamped visual record, full command and sensor log, and operator identity. The package format has been shown to a tower operator and confirmed as sufficient for their compliance file.
Project 20. Witnessed intervention on a real tower
Objective
Perform a complete teleoperated maintenance intervention on an actual telecommunications tower, witnessed by a tower operator, with no human on the structure.
Why it matters
This is the artefact that converts a project into a company. Everything before it is preparation.
Done when
The intervention completes on real steel. A tower operator representative witnesses it and provides written feedback. The evidence package is delivered and assessed. Recorded in full for use in commercial conversations.
Sequencing notes
What can run in parallel. Phase 3 perception work is largely independent of Phases 1, 2 and 4 and can proceed alongside them with a second person or in alternating blocks. Everything else is dependent.
What cannot be skipped. Project 4 gates all climbing work. Projects 16 and 17 gate all work on live infrastructure. These are not sequencing preferences; they are the conditions under which anyone will let you on a tower.
On pace. Mechanical iteration has physical cycle times. Parts arrive when they arrive; prints take as long as they take; a redesigned bracket costs a week regardless of hours worked. Effort compresses the software and the analysis. It does not compress the machining. Plan the mechanical path first and fill the waiting with software, perception and documentation.
On documentation. Every project produces a written record: what was built, what was measured, what failed, what was changed. This is not administrative overhead — it is the raw material of a patent application, a safety case and a technical due diligence, all three of which arrive sooner than expected.
What runs alongside the ladder
The ladder is the engineering programme. It is necessary and not sufficient. Three things must progress in parallel, and they determine whether the engineering ever reaches a customer:
Site access. Written permission to develop on a real tower. Without it the programme stops at Project 14. This is the first thing to secure and the hardest to substitute.
Safety competence. A work-at-height and structural specialist engaged early, shaping the design rather than assessing it afterwards.
Operator conversations. What tasks they would pay to have done remotely, what accuracy they require, what evidence they would accept, and what would have to be true for them to permit a machine on their structure.
The engineering answers whether the machine can work. These answer whether it matters.