Generative Routing Topologies for EV Harnesses
The wiring harness is the third heaviest component in an electric vehicle, behind the chassis and the battery pack. Optimizing its routing is no longer a human-scale problem.
The Multivariable Routing Problem
When an engineer manually routes a harness in CAD (like CATIA or NX), they rely on heuristics: keep away from hot zones, follow the sheet metal, bundle things together to save clips. But in EVs, the constraints are exponentially more complex:
- High Voltage vs Low Voltage EMI: You cannot run 400V/800V lines parallel to sensitive CAN bus networks without massive shielding, which adds weight.
- Thermal Derating: Running a harness near the inverter means the ambient temperature is higher, so you must use a larger wire gauge to carry the same current safely. Larger gauge = more copper = more weight.
- Manufacturability: A mathematically perfect route might require an assembly line worker to bend their wrist at an impossible angle to install a clip.
| Routing Method | Average Copper Weight (EV) | Iteration Speed |
|---|---|---|
| Manual/Heuristic | ~32 kg | Weeks |
| Generative AI (Zonal) | ~26 kg | Hours |
How Generative AI Solves It
Generative design tools take the 3D space of the vehicle and a netlist (what needs to connect to what). You input the constraints (max temperature zones, keep-out zones, EMI clearance rules). The AI then generates thousands of potential topologies.
It uses genetic algorithms to "evolve" the harness. It might find that routing a bundle longer to stay in a cooler zone allows for a smaller wire gauge, resulting in a net weight reduction despite the longer physical path.
The Zonal Architecture Shift
The output of generative routing is pushing the industry away from traditional domain architectures (where wires run from a central ECU to every sensor) towards zonal architectures. AI models consistently prove that grouping compute by physical zone (e.g., a left-front controller) and networking the zones together uses drastically less wire.