Soorya Boopal

About

Robotics software engineer with ~4 years of ROS experience who implemented a real-time diffusion-policy control loop entirely in Gazebo, focusing on inference-latency reduction (warm-start + truncated denoising) for stable closed-loop execution. Has hands-on experience building custom ROS control nodes, optimizing AMR navigation (SLAM + RRT) with sensor-fusion for dynamic obstacles, and designing deterministic multi-robot coordination; also uses Dockerized ROS environments and automated simulation/benchmark pipelines.

Experience

Education

Key Strengths

• Reduced diffusion-policy inference latency for real-time closed-loop control via warm-starting and truncated denoising
• Stabilized simulation-to-control execution by enforcing fixed control frequency and decoupling policy inference from Gazebo physics steps
• Mitigated distribution shift with action smoothing and safety bounds to prevent error accumulation
• Built custom ROS control node subscribing to robot state and publishing joint commands at fixed frequency with action constraints
• Optimized AMR navigation in Gazebo using ROS navigation stack (SLAM + RRT planning) under dynamic obstacles
• Designed rule-based, confidence-weighted sensor fusion (LiDAR/depth/ultrasonic) to improve obstacle estimation and safety
• Implemented centralized multi-robot coordination with timestamping/sequence checks and non-blocking latest-state snapshots for real-time safety
• Improved reproducibility and deployment reliability using Dockerized ROS environments and scripted simulation/benchmark pipelines