01Why the joint decides
Software reuse works because interfaces are cheap: a function call costs nanoseconds and weighs nothing. A physical interface is expensive in every currency that matters to a machine: mass, stiffness, volume, contact resistance, latency and failure probability. Industrial tool changers, modular robot research platforms and reconfigurable fixtures all show the same pattern: the module is the easy part, and the connection determines what the system can do.
The modular approach KyperX is researching goes a step further by letting modules carry structural load as part of the machine’s frame. That turns the interface from an attachment into a load path, which raises the bar for every requirement below.
- Module
Connects through
- Mount
- Power
- Data
Functions a module can carry
- Sense
- Compute
- Comms
- Actuate
- Optional backplane
02Mechanical: repeatability, preload and stiffness
A structural interface has to locate a module to the same position every time, carry static and dynamic loads, and do so through thousands of mating cycles. Kinematic or quasi-kinematic coupling, with three well-defined contact pairs, gives very high repeatability without over-constraint; adding preload makes the joint stiff, but preload has to be generated and released by something, which is where retention mechanisms, actuators and failure modes arrive.
Stiffness is not a structural nicety. Every compliant joint in a kinematic chain lowers the first resonant frequency of the machine, and that frequency caps the achievable control bandwidth. A modular arm with several soft interfaces will be slower and less accurate than a monolithic one, however good its motors.
| Choice | Gains | Costs |
|---|---|---|
| Kinematic coupling with preload | Repeatability, defined load path | Retention mechanism, preload actuation |
| Dovetail or rail | Simple, tolerant of misalignment | Lower stiffness, wear, play over cycles |
| Bolted flange | Highest stiffness | Tools, time, no fast reconfiguration |
| Magnetic retention | Fast mating, self-aligning | Limited moment capacity, stray fields |
03Electrical: power through a connector
Power has to be delivered to the most demanding module without penalising the simplest. Bus voltage is the first decision: a higher extra-low voltage keeps conductor currents manageable for motor modules while staying inside the safety extra-low-voltage band that simplifies protection design; lower voltages push current, contact count and voltage drop up quickly.
Three connector problems recur. Contact resistance rises with wear and contamination, and at high current a few milliohms becomes heat. Hot-plugging creates inrush and arcing unless contacts are sequenced and loads soft-started. And regenerative energy from decelerating actuators has to go somewhere, so a shared bus needs a strategy for absorbing it.
04Data: bandwidth is not the hard part
Most module data rates are modest. The hard requirements are determinism and time. A servo loop needs its data on time every cycle, with jitter well below the period, and coordinated motion across modules needs a shared clock. Field-bus choices such as CAN FD, EtherCAT and time-sensitive Ethernet differ less in raw bandwidth than in latency guarantees, topology, synchronisation and how many contacts they consume in the connector.
A reusable module also has to identify itself. Discovery, configuration and versioning sound like software details, but they decide whether a machine can be assembled from modules without hand-written integration code each time.
05Thermal, EMC and safety
The interface is also a thermal path, often the only one between a hot module and the structure that could sink its heat, and an electromagnetic boundary where shielding and grounding either continue or break. Both are routinely left to late design stages, when the connector is already fixed.
Safety cannot be added per module. Emergency stop, safe torque off and safety-rated monitoring have to run through the interface as a continuous, verifiable chain, and the architecture has to support the performance level the target machines require under standards such as ISO 13849. A modular system is only as safe as its least verifiable joint.
06Evaluating an interface before building it
Each requirement can be turned into a measurable criterion and checked against candidate designs on paper, in simulation and on simple rigs before any module is committed. The point is to make the trade-offs explicit and comparable.
| Requirement | Criterion | Early evidence |
|---|---|---|
| Location | Positional repeatability over rated mating cycles | Tolerance stack-up; cycle test on a coupling rig |
| Load path | Joint stiffness and moment capacity per axis | Finite-element model; static load test |
| Dynamics | Effect on the first structural mode of target machines | Modal model of representative assemblies |
| Power | Continuous and peak current per contact, voltage drop, temperature rise | Connector rating; thermal test at peak duty |
| Timing | Cycle time and jitter at the target loop rate | Bus timing analysis; bench network test |
| Safety | Achievable performance level of the stop and monitoring chain | Safety architecture review against the target standard |
07The parameters that lock in early
Several decisions become irreversible once modules exist: geometry and footprint, the connection standard, bandwidth and timing guarantees, and the retention method. These are the open engineering parameters in KyperX’s current work, and they are deliberately being defined through requirements and evaluation criteria before any hardware commitment.