ORB-SLAM3 INTEGRATION FOR UAV
Monocular + inertial SLAM ported to a custom ROS2 driver on an ArduPilot quad, enabling GPS-denied state estimation in indoor warehouses.
> MOUNTING SENSOR ARRAY ................ OK
> CALIBRATING IMU ..................... OK
> SPINNING UP LIDAR .................... OK
> LOADING ROS2 WORKSPACE ............... OK
> STARTING SLAM PIPELINE ............... OK
> ESTABLISHING TELEMETRY LINK .......... OK
> POWERING DRIVE MOTORS ............... OK
SYSTEM ONLINE █
< AUTONOMOUS SYSTEMS ENGINEER />
> ▊
// SYSTEM PROFILE
root@robot:~$ ./init --role roboticist
> Loading profile ... done
> NAME : HARSH BHARGAV
> FOCUS : SLAM / PATH PLANNING / CONTROL
> STACK : ROS2 · C++ · PYTHON · GAZEBO
> MODALITY : PERCEPTION → NAV → ACTUATION
> STATUS : READY FOR DEPLOYMENT
I design autonomous systems that perceive their world, plan
their path, and move with intent. From indoor SLAM pipelines
to outdoor GPS-denied navigation on ArduPilot UAVs, I turn
sensor streams into decisions.
> PRESS_SKILL_MATRIX ▮
I bridge the gap between raw silicon and autonomous behavior — writing drivers, fusing sensors, solving the localization problem, and closing the loop with motion control.
Vision + LiDAR + IMU fusion. SLAM, VIO, and semantic understanding that map the world into a model the robot can trust.
A*, RRT*, and model-predictive controllers that choose paths under uncertainty — fast, safe, and real-time.
PID and cascaded controllers on real hardware — brushed motors, ESCs, servo joints — with safety limits that never quit.
// CAPABILITY REPORT
// DEPLOYED PAYLOADS
Monocular + inertial SLAM ported to a custom ROS2 driver on an ArduPilot quad, enabling GPS-denied state estimation in indoor warehouses.
Full-stack outdoor robot running ROS2 Nav2 — costmap-based global and local planners, recovery behaviors, and EKF-fused odometry for 4-hour missions.
Kalibr-calibrated camera–IMU rig with tightly-coupled VIO, evaluated against ground truth on TUM/EuRoC-style trajectories for drift characterization.
// LIVE KINEMATICS SIMULATOR
> Drag sliders to command joints
// DEPLOYMENT HISTORY
Building visual-inertial SLAM systems on ArduPilot UAVs; benchmarked drift at 0.6% over 1.2 km indoor loops.
Shipped Nav2-based navigation stack for an autonomous mower; cut mapping time 40% via optimized costmap layers.
Designed PID controllers and telemetry dashboards for 6-DOF manipulator arms; reduced settling time by 35%.
Led a 6-member team to a national win; designed swerve-drive kinematics and vision-based gate detection.
// ESTABLISH CONTACT