Columbia University | Robotics Research Project
Project Overview
Developed an advanced Evolutionary Algorithm (EA) to autonomously optimize the morphology and gait of simulated soft robots. The project involved engineering a custom physics engine for high-fidelity performance modeling and integrating cloud-scale computing to execute massive iterative searches for optimal locomotion.
Technical Responsibilities & Methodologies
Evolutionary Algorithm Design: Developed and implemented a sophisticated EA using genetic operators (mutation and crossover) to optimize the physical control parameters of non-rigid structures.
Custom Physics Engine Engineering: Built a physics engine from the ground up to realistically simulate the complex, dynamic movements of soft bodies, ensuring reliable performance evaluation during evolution.
Cloud Infrastructure Integration: Leveraged the Google Cloud Platform (GCP) to scale computational resources, enabling the algorithm to process over 100,000 iterations with high efficiency.
Data Analytics & Validation: Managed and analyzed massive simulation datasets to track evolutionary progress and validate the accuracy of the custom-built modeling engine.
Key Achievements
Performance Milestone: Successfully evolved a soft robotic structure that reached a significant walking speed milestone of 0.6 m/s.
Computational Scalability: Achieved deep evolutionary search capabilities by utilizing cloud resources, significantly outperforming local computing constraints.
Scientific Contribution: Demonstrated the potential for automatic design and optimization of compliant, non-rigid structures through evolutionary computation.