What did I do this week?

This week, I worked further on resolving Issue #10, which involves replacing the legacy qutip_qtrl.fid_computer with a modern and modular FidelityComputer class designed for quantum optimal control workflows.

Previously, I had submitted a preliminary version of this class supporting multiple fidelity types (PSU, SU, TRACEDIFF) and ensuring compatibility with JAX for automatic differentiation. However, based on further reflection, I decided to hold off on raising a pull request and instead focused on polishing the implementation.

The major highlight of the week was addressing the previously failing test cases for:

State-to-state GRAPE optimization
State-to-state CRAB optimization
Parameterized CRAB optimization

These failures were caused by subtle differences in fidelity expectations and numerical tolerances when switching to the new engine. I resolved them by:

Ensuring consistent handling of input shapes and data types
Carefully aligning fidelity formulae with the expected definitions in the test suite
Improving numerical precision and tolerances where required

As a result, all tests now pass, including the edge cases that initially failed.

Beyond that, the direction I’ve been shaping for the FidelityComputer revolves around:

Modular structure: With clearer separation of state, process, and map fidelities in mind, the aim is to make each component independently extensible and easier to maintain.
Batch support: Handling multiple quantum states or unitaries efficiently has been central to the design approach, and continues to shape how I think about scaling and performance.
Cleaner integration: Ensuring compatibility with GRAPE, CRAB, and other optimization methods remains a key goal — something I’ve been aligning the implementation toward as the project evolves.

While the code is functionally complete, I plan to further refine and customize certain parts before raising the pull request.

Plan for next week?

Next week, I’ll focus on:

Studying the relevant sections from Alex’s PhD thesis — especially Section 1.6: Quantum control function optimisation — to deepen my understanding of GRAPE
Finalizing and cleaning up the FidelityComputer interface
Contributing to documentation improvements as part of Issue #13, particularly:
- Investigating the broken search functionality on the QuTiP-QOC documentation site
Raising a well-documented and review-ready pull request

I’m excited to be moving closer to a stable and efficient fidelity framework that can serve as the backbone of QuTiP-QOC’s optimization tools.