Changda Tian

I am a robotics researcher and PhD fellow of the Marie Skłodowska-Curie RAICAM network at FORTH in Heraklion, Greece, affiliated with the University of Crete.

Now

My focus is modular legged robot design and control: I build Modbot, a reconfigurable biped module that can link into a quadruped, and train configuration-conditioned reinforcement-learning controllers that serve the whole robot family. The goal is to measure what modular flexibility actually costs — in reliability, control, and energy — within RAICAM's mission of mobile robots for inspecting and maintaining critical infrastructure.

Before

I trained in Automation and Control Engineering at Shanghai Jiao Tong University (Zhiyuan College Honors Program), and did my master's in the RL2 Lab of the AI Institute with Prof. Yue Gao, working on legged robots end to end:

• Capability-aware locomotion — taught the hexapod Qingzhui to know which terrain it can traverse, under which controller and topology, and to plan long-range paths with that knowledge. Validated in simulation and on the real robot (ROBIO 2022 Best Paper in Biomimetics finalist; extended at RCAR 2024).
• Adversarial balance control — stance legs on whole-body control, swing legs on reinforcement learning, trained against each other for fast, stable velocity tracking.
• HMAMP — human-style tool manipulation from adversarial motion priors, evaluated on hammering and deployed on a real Kinova Gen3 arm (Robotica, 2025).
• Winter Olympics robots — with Prof. Feng Gao's team, converted hexapods to skate and ski and field-tested them at real rinks and resorts for the Beijing 2022 torch relay.

Earlier still: my first sim-to-real hexapod controller as an undergraduate (IEEE CYBER 2019), and a summer at Caltech's Moore Lab implementing neural-network primitives on FPGAs.

Collaboration

I'm glad to talk about modular robots, locomotion learning, and open-source robotics — email is the fastest channel.