Skip to content
ranbao_deng://

~/projects

Projects

Things I have built, from wearable legged robots and quadrotor planners to learning pipelines for haptics and AI for a strategy game. Cards marked → have a case study, and ↗ links to code or an external page. Team projects are labelled as such.

currently · Sep 2026 – present

Student Developer (part-time) at X-Frontiers ↗

Part-time (~16 h/week) developer at a student-run tech company, helping to bring an AI-driven rescue drone from prototype to field deployment.

research2026

Machine learning for electrovibration haptics

Predicting finger–touchscreen friction under electrostatic actuation from small, highly user-dependent data. The work compares six model families, tests cross-participant generalisation, and identifies a real-time-feasible feature set.

friction-force prediction, random split (TabPFN)
R² 0.91
on unseen participants: the open problem
R² ≈ 0.5
participants × trials, 6 model families
10 × 3
  • Python
  • scikit-learn
  • TabPFN
  • Haptics

engineering · BEng thesis2021

“Centaur” load-carrying walking robot

A pair of robotic legs that walks with its wearer and carries the load. I built the IMU-based gait perception and the phase-synchronised gait control. They were tested in Simulink and then on a 15 kg prototype during 10-minute walking trials.

  • MATLAB/Simulink
  • IMU sensor fusion
  • Legged robots
  • Control

engineering · Team project2022

MPC ball-catching quadrotor (opens in a new tab)

A linear model predictive controller on linearised quadrotor dynamics, with a DARE-based terminal cost to guarantee stability and recursive feasibility. Validated in 3D simulation across different horizons and constraint sets.

  • Python
  • MPC
  • Control

engineering · Team project2022

k-PRM + A* quadrotor motion planner (opens in a new tab)

A ROS planner built from scratch: occupancy mapping, k-PRM with a KD-tree, A* search, and minimum-snap and corridor-based trajectory optimisation (solved with OSQP). Benchmarked against RRT variants in forest and maze environments.

  • ROS
  • C++
  • Motion planning
  • Trajectory optimisation

engineering · Solo2026

AI Cup 2026: bird species from radar

A solo entry in a national challenge run by Team Epoch and TNO: an ensemble classifier for bird species from radar tracks, plus a system design that turns calibrated predictions into targeted wind-turbine curtailment.

teams on the final hidden test set, as a one-person team
26 / 89
  • Python
  • Classification
  • Kaggle

side project2026

AI Build Cities: Crusader Kings III mod

In the base game, AI rulers build almost nothing but castles. This mod rewrites their holding-building logic so that castles, cities and temples are chosen by terrain, culture, government type and ruler personality. It's a small, focused behaviour change, and players adopted it widely: #14 in the Workshop's most-popular list for the past year and #10 for the past six months (Sep 2026), five months after its April release.

subscribers
14,400+
unique visitors
37,500+
ratings, overwhelmingly positive
178
  • Game AI
  • Paradox script
  • Steam Workshop

research · Research project2020

Elastic backpack with electrostatic damping

An elastically suspended back frame that reduces the load on the carrier. I designed and prototyped the electrostatic damping mechanism and ran the motion-capture experiments, which led to an IEEE ICARM 2021 paper and a Chinese patent.

  • Mechanism design
  • Motion capture
  • Prototyping

engineering · Internship2019

Terramechanics simulation for tractors

A research internship at the University of Notre Dame (USA): Webots simulation of tractor performance on soft soils using terramechanics models, supporting AgJunction's wheel-navigation products.

  • Webots
  • Simulation
  • Vehicle dynamics

Looking for the research side? The electrovibration work, thesis and literature review are covered in detail on the research page.