This is a refreshingly concrete autonomy claim. Toyota Research Institute's grant US11958498B2, "Data-driven warm start selection for optimization-based trajectory planning" (issued April 16, 2024), targets a specific, well-understood computational problem: optimization-based planners converge faster when seeded with a good initial guess, and a bad start wastes compute or fails to converge in time. The claim is about selecting that warm start from learned data. The CPC pairs AV-control (B60W 60/001, G05D 1/0212) with machine-learning classes (G06N 3/08, G06N 20/00).
The enabling content is unambiguous because the problem is well-defined. A trajectory optimizer is a known component; warm-starting it is a known speedup technique; using learned data to pick the warm start is the claimed novelty. That is an engineering improvement with a measurable effect — faster convergence, more reliable real-time planning — which is exactly what an enabling claim should recite. There is no autonomy-level hand-waving here; it is a planner-internals optimization.
On scope, the independent claim establishes the data-driven warm-start selection; the dependents that specify how the learned model is trained and how the start is chosen are the moat. Because this is a granted B2, the scope has survived examination, and the planning-optimization framing keeps it tighter than a broad behavior claim. The defensible piece is the selection method, and prior-art tension with general warm-start techniques is the obvious examination hurdle it cleared.
Strategically, Toyota Research Institute filing planner-internals IP — rather than splashy full-autonomy claims — fits its research-lab identity and the broader 2024 maturation of the field toward optimizing known components. Dated April 2024, it reflects an industry past the demo phase and into the engineering of latency and reliability. The verdict: clearly enabling, narrowly scoped, and issued — the kind of quiet, defensible planning patent that does real work. Read the training-method dependents for the specifics.