Learning in any of these models can take three general forms:
- The edge devices can each independently learn, and share their learning through the hub (108)
- The edge devices send raw data to the hub (108), which learns from the data and sends behavioral updates to the edge devices (the hub 108 does the learning)
- Some combination of learning takes place on the edge devices and the hub (108). As an example, each edge device could perform some level of unsupervised learning based exclusively on its own inputs, while the hub (108) performs supervised learning across the inputs from all the edge devices and shares this learning globally.
The odds are good that your IoT invention utilizes some form of unsupervised, semi-supervised, or supervised learning, or a combination of these. Obtaining patent protection, or even if patent protection should be sought, will depend to some extent on the approach you take to IoT learning.
For example, if your approach is of the second kind (learning is centralized in the hub), your invention might be a candidate for trade secret protection. This is because it might be difficult for a competitor to independently develop your learning algorithms, or to discover them from the IoT interface to your hub. In that case, you have to weigh the tradeoffs of trade secret protection against the costs and benefits of patent protection.
If your invention uses learning of type (1) or (3), your algorithms might be easily discovered through reverse engineering. If they are valuable to your competitive advantage, then patent protection is probably your best bet for preventing competitors from taking your algorithms and using them in their own products.
Most hybrid mesh (200) or full mesh (300) IoT models will use some variant of type (1) or (3) learning, so they are strong candidates for patent protection. That’s not to say patent protection doesn’t make sense for other IoT models – it certainly does, but the option for trade secret protection is also there when the primary learning takes place at the hub and isn’t easily discovered.