Portfolio

RoboDrive Competition (ICRA 2024)

Participated in the RoboDrive Challenge, focusing on robust multi-modal BEV 3D object detection, using corrupted camera and LiDAR data from the NuScenes dataset. Achieved 4th place with scores of 42.8 (NDS) and 26.4 (mAP), and got invited to present at ICRA 2024 conference.

Training Diffusion Models using Reinforcement Learning

Under the guidance of Prof. Biplab Banerjee, focusing on optimizing diffusion models for downstream tasks using reinforcement learning (RL).

Proximal Policy Optimization for Cart-Pole Balancing

Implemented Proximal Policy Optimization (PPO) for CartPole Balancing with Generalized Advantage Estimation (GAE) and clip loss.

Thivyanth Venkateswaran

Portfolio

RoboDrive Competition (ICRA 2024)

Training Diffusion Models using Reinforcement Learning

Proximal Policy Optimization for Cart-Pole Balancing