Research

Exploring the intersection of ML and robotics

Published Research

Machine Learning-Based PID Auto-Tuning for Robotic Systems

Makhatov, I. (2025)

TechRxiv (IEEE)

Abstract

This paper presents a machine learning approach to automatically tune PID controllers for robotic systems. Traditional manual tuning requires expert knowledge and can take hours per robot. Our ML model predicts optimal PID parameters in less than 0.2ms, achieving 78-90% performance improvement over manual tuning methods. The model was trained on 1000 robot configurations in simulation and demonstrates strong generalization across different robot types and control scenarios.

Research Interests

Primary Areas
  • Machine Learning for Control Systems
  • Robotics & Mechatronics
  • Human-Robot Interaction
Specific Topics
  • Adaptive control algorithms
  • Transfer learning for robotics
  • Assistive technologies
  • Edge AI for embedded systems

Summer Programs

nFactorial Incubator
Summer Incubator Program | Jun 2025
Selectivity: 1 of 75 from 2,000+ international applicants (Top ~4%)
Grant: Chevron grant offer
ISSAI — Nazarbayev University
Summer Research Program | Jun 2024
Selectivity: 1 of 30 from ~350 international applicants (Top ~9%)
Focus: Month-long team project on tensegrity structures
Skills: CAD modeling and AI methods; mentored seminars
KAIST IT Training Course
Jul 2025
Selectivity: 1 of 80 from across 20+ NIS schools
Two-week International intensive: AI, programming, design coursework
Led team to 1st place in AI/programming track competition

What's Next

Current

Hardware validation for ML PID paper

Testing on physical robots (planned Q1 2026)

Interested in

  • Reinforcement learning for manipulation
  • Sim-to-real transfer
  • Robotic rehabilitation systems

Open to

  • Research internships
  • Collaboration opportunities
  • Graduate programs (2026+)