Enabling or Aspirational? GM's 2020 Grant on Automatic Rule-Learning for Self-Driving

Enabling claim or aspiration? That is always the first question for an autonomy patent, and GM's grant US10860022B2, "Method and apparatus for automatical rule learning for autonomous driving" (issued December 8, 2020), sits right on the line. The independent claim describes deriving driving rules automatically — learning the policy rather than writing it by hand. The CPC is sparse and pure: G05D 1/0088 and G05D 1/0221, both autonomous-control classes, with none of the sensor-hardware classes that would anchor it to a specific perception modality.

That sparseness is a signal. A patent claiming a learning method without naming the sensor stack or the operational design domain is claiming the abstraction, not the deployment. The enabling content is the rule-learning mechanism itself; what it does not do is guarantee a capability in a real vehicle, because the hard parts — what data, what generalization, what safety envelope — live outside the claim. Read it as positioning on the methodology, not evidence of a shipped autonomous behavior.

Still, the strategic logic is sound. By late 2020 the industry consensus had shifted toward learned behavior policies over rule books, because hand-authored rules do not cover the long tail of real driving. GM filing on automatic rule-learning stakes a claim at the layer everyone was migrating toward. For an IP correspondent, that is a sensible place to hold a position even if the claim's breadth makes it more of a defensive marker than a narrow, easily-enforced grant.

The skeptic's note: do not treat this filing as a deployed capability. The grant tells you GM was investing in the learned-policy layer in 2020, and the absence of sensor and ODD limitations tells you the scope is conceptual. The dependent claims would have to add concrete training or validation steps to harden it. As an enabling-versus-aspirational call, this one leans aspirational in framing but enabling in the specific method it recites — which is exactly why the independent claim, not the abstract, is the thing to read.