This project centered on the development and application of an advanced Evolutionary Algorithm (EA) to significantly enhance the walking capabilities and performance of simulated soft robots. The core innovation involved engineering a custom-built physics engine for realistic performance modeling and integrating cloud resources to handle the extensive computational demands of the optimization process.
Developed and implemented a sophisticated Evolutionary Algorithm, utilizing core genetic operators such as mutation and crossover techniques, specifically tailored to optimize the physical and control parameters of soft robot designs.
Engineered a Custom Physics Engine from scratch to accurately and realistically simulate the complex, dynamic movements of the evolving soft robots, which was crucial for reliable performance evaluation.
Integrated Google Cloud Platform (GCP) to scale computational resources, significantly optimizing the algorithm's efficiency and enabling the execution of rapid, large-scale evolution iterations.
Managed and analyzed the simulation data across over 100,000 iterations to track performance improvement and validate the success of the evolutionary process.
Successfully demonstrated the power of the algorithm by reaching a significant performance milestone: an evolved soft robot walking speed of $0.6 \text{ m/s}$.
Validated the precision and accuracy of the custom-engineered physics engine through consistent and reliable performance modeling results.
Achieved high computational efficiency and scalability by integrating cloud resources, enabling faster iterations and deeper evolutionary search compared to local computing.
Contributed to the field of robotics by showcasing the potential of evolutionary computation for the automatic design and optimization of compliant, non-rigid structures.