Ardalan Aryashad

Ph.D. Student, Aerospace & Mechanical Engineering · USC
M.S. Student, Department of Computer Science · USC

I’m a third-year AME Ph.D. student at USC, advised by Prof. Yan Jin. I have recently started an M.S. in Computer Science.

At the IMPACT Lab, I am currently working on vision-language-action (VLA) models and their applications in robotic task planning and manipulation, and agentic AI approaches for MBSE modeler tasks.

Email  ·  CV  ·  Scholar  ·  LinkedIn  ·  GitHub

prof_pic.webp

news

May 18, 2026 SEMAA accepted to ASME IDETC-CIE 2026.
May 13, 2026 FSAE 2026: USC Racing won Autocross (1st), placed 7th overall, and reached the Design Finals.
Mar 06, 2026 Filters to VLMs accepted as an Oral and presented at WACV 2026, Tucson, AZ (Mar 6–10).
Aug 25, 2025 Started my MS in Computer Science at USC.
Aug 17, 2025 VLAG presented at ASME IDETC/CIE 2025, Anaheim, CA (Aug 17–20).
May 19, 2025 VLAG accepted to ASME IDETC/CIE 2025.
May 14, 2025 FSAE 2025 : USC Racing’s first-ever podium — 3rd in Autocross and a Design Finals appearance.
Aug 21, 2023 Started my PhD in Mechanical Engineering at USC.

publications

  1. WACV 2026
    fog.webp
    From Filters to VLMs: Benchmarking Defogging Methods through Object Detection and Segmentation Performance
    Ardalan Aryashad, Parsa Razmara, Amin Mahjoub, Seyedarmin Azizi, Mahdi Salmani, and Arad Firouzkouhi
    In IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshops, 2026
    arXiv Website
  2. IDETC 2025
    vlag.webp
    VLAG: Graph-Based Planning for Vision-Language-Action Models
    Ardalan Aryashad, and Yan Jin
    In ASME International Design Engineering Technical Conferences (IDETC/CIE), 2025
    DOI Website

projects

Fall 2024 – Present
USC Racing — Suspension & Data Acquisition
USC Racing — Suspension & Data Acquisition
Two years with USC Racing, split between hands-on suspension work and building tools that made the car’s data useful at the track.
Website
2025–Ongoing
SEMAA: Agentic AI for Model-Based Systems Engineering
SEMAA: Agentic AI for Model-Based Systems Engineering
SEMAA brings conversational AI into everyday systems-engineering work, from editing diagrams to working with private engineering data.
Website
Spring 2026
Merging Robot Brains (and Why It Didn't Work)
Merging Robot Brains (and Why It Didn't Work)
An experiment in merging task-specific robot policies that produced a clear negative result: averaging their weights did not preserve the skills.
Website
Fall 2024
Walking Biped Robot
Walking Biped Robot
We used a single model-predictive controller to make a simulated biped stand, walk, run 10 meters, and climb stairs.
Website
Fall 2023
Automotive Predictive Maintenance
Automotive Predictive Maintenance
A course project that took messy automotive sensor data from preprocessing and model comparison to a working maintenance app.
Website