Research Portfolio

I currently work on reach-avoid control for stochastic systems

The core question is straightforward: once the initial set, target set, unsafe set, and probability requirement are fixed, how can we actually produce a controller with a formal guarantee instead of a policy that only looks good in simulation? My current pipeline has four steps: train a reference controller with reinforcement learning, approximate it with a PAC-guided polynomial controller, synthesize a stochastic barrier-like certificate through SOS / SDP, and then continue with controller or certificate iteration when needed.

Top 1Best overall score in the 2024 graduate entrance exam for the target major
3.95Undergraduate degree-course GPA; the transcript is now included on this site
2nd · 3rd · HMHuawei Cup second prize, third prize, and MCM/ICM Honourable Mention
What I am doing now

My current research workflow

This is neither pure controller design nor pure verification. What matters to me now is keeping policy learning, verifiable approximation, and probability certification inside one connected pipeline.

Formulate the reach-avoid task clearly

I study infinite-horizon reach-avoid control for stochastic dynamical systems under fixed initial, target, unsafe, and probability requirements.

Turn the learned policy into something verifiable

A SAC controller is first trained as the reference policy, and PAC approximation is then used to produce a polynomial controller for formal reasoning.

Read probability, time, and iteration together

The output is not only a certified lower bound; I also keep track of timing, controller degree, certificate degree, and where to iterate next.

Research focus

The homepage now stays at the overview level

  • This page now keeps only a concise summary of what I work on, what problem I am trying to solve, and what the main method chain looks like.
  • More detailed technical material has been moved off the homepage and into the research and resume pages.
  • This keeps the homepage lighter as a public-facing entry point, while the detailed technical material remains on the research and resume pages.
Quick navigation

Where to continue

  • Research: full problem setup, method pipeline, paper, and experiment results.
  • Profile: academic history, awards, proof photos, and growth path.
  • Projects: open-source repositories, project context, award images, and engineering practice.
Selected materials

Material entry points already organized on the site

Transcript

Undergraduate transcript

The transcript has been added to both the profile page and the resume page for GPA, course structure, and credits.

Transcript preview
Evidence

Proof photos and certificates

Graduate and undergraduate awards, volunteer certificates, training certificates, and activity proofs are organized on the profile and resume pages.

Huawei Cup certificate preview
Projects

Projects and repositories

The projects page now includes the graduation project, the internship project, their GitHub repositories, and supporting award images.

Project-related image preview