Projects

Quality-Diversity Reinforcement Learning for Damage Recovery in Robotics

A QD reinforcement learning-based research to enable robots to recover from unexpected mechanical damages in a handful of minutes.

ResearchRoboticsReinforcement LearningJAXIn Progress

Bidding Hub

A microservice-based real-time auction platform to browse live auctions, place competitive bids, and manage the listings.

SoftwareFlaskDockerReactPostgreSQLWebsockets

Driving Condition-based Energy Management Strategy of Hybrid Vehicles

Developed an online driving condition recognition algorithm using Time Convolutional Networks (TCN) to accurately classify real-time driving conditions. Designed a condition-adaptive energy management strategy, demonstrated lower overall energy consumption across varying driving scenarios.

ResearchReinforcement LearningDDPGTCNPytorch

Variational Autoencoder-based Driving Condition Generation

Developed a driving cycle generation algorithm using a Variational Autoencoder (VAE) to create a representative driving condition database, enabling a lower-cost alternative to real-world data collection while maintaining accuracy and diversity in driving conditions.

ResearchVariational AutoencoderLSTMPCA