Kartik Virmani

Built a full-stack ROS 2 autonomy pipeline integrating navigation, SLAM, motion planning (RRT, Pure Pursuit, MPC), and perception with LiDAR/IMU/RealSense.

Implemented 3D path planning, vision-based localization, and PID/MPC controllers in a custom simulator and hardware-in-the-loop tests.

Integrated GroundingDINO, SAM, and Kalman Filters in ORB-SLAM3 for dynamic object masking, reducing ATE by 29.5% and boosting map density by 81%.

Streamed live RGB-D data, ran ORB-SLAM3 for dense indoor mapping, and visualized pose + occupancy in real time with Intel RealSense.

Implemented Stein Variational Guided MPPI on F1/10 platform, reducing track error by 38.7 m and cross-track error to 0.855 m at 50 Hz.

Built random-shooting MPC and PPO agents for Cartpole, Walker, and Humanoid in MuJoCo, achieving 700+ avg returns and smooth convergence.

Fused IMU, GPS, LiDAR, and encoder data via Unscented Kalman Filter to reduce localization drift by 30% over 1 km.

Implemented NeRF in PyTorch for novel-view synthesis with positional encoding, hierarchical sampling, and PSNR/SSIM metrics.

Combined neural networks with MPC in CarSim+MATLAB, cutting planning convergence by 30% with CUDA acceleration.

Designed MPC with Zero Momentum Control for quadruped locomotion in PyDrake/MuJoCo, boosting planning robustness by 8%.

C++ autonomous vehicle arena game with 90% task success, multi-agent ROS simulation, and collision-avoidance testing.

Built ESP32 modular mobile robot with TOF sensors, H-Bridge drivers, PID encoder loops, and WiFi-based multi-ESP communication.

Implemented PID-based roll correction that detects >5° roll at ≥31 mph and adjusts steering, reducing rollover risk by 30%.

Simulated thermal barrier coating cracks under cyclic heat loads in ABAQUS XFEM, optimizing geometry for durability.

Ran transient RANS simulations in STAR-CCM+ to optimize undertray & diffuser aerodynamics across pitch angles.

Designed rear-wheel steerable suspension in SolidWorks & ANSYS, validating safety under ≤700 N with GD&T-defined tolerances.

Modeled heat generation & wear in ANSYS, optimizing contact geometry and material choice to reduce fatigue.