Shishir Kolathaya

Optimizing actuator parameters alongside mechanical design and control yields a 42% jump distance improvement in a five-bar monoped robot, demonstrating the significant impact of holistic co-design.

Lagrangian Neural Networks embedded within an RL MPC framework enable quadrupedal robots to achieve 4x faster planning with minimal performance loss by avoiding costly matrix inversions.

Legged robots can now nimbly navigate complex terrains by fusing learned visual perception with infinite-horizon planning, achieving robustness beyond traditional MPC or RL approaches.

By disentangling rigid-body mechanics from stochastic interaction effects, STRIDE achieves more accurate and reliable dynamics prediction for robots operating in uncertain environments.