My research develops safety-critical control and motion planning for robots and multi-agent systems, pairing rigorous control theory with hands-on validation on real hardware. The projects below are ordered from my current work back to my earliest research.

Table of Contents

• Safety-Critical Control — Current Research
• Autonomous Vehicle Planning and Control
  • Clothoid-Based Path Planning
  • Non-linear Rear-Wheel Feedback Control
  • Vehicle Dynamics with Cruise and Lateral Control
  • Pure-Pursuit Control
• Autonomous Quadrotor UAV Control
  • Cascaded PID Control in PX4
  • State-Dependent LQR for Trajectory Tracking
  • Offboard Velocity Control
  • Minimum-Snap 3D Trajectory Generation
• Distributed Control, Coordination, and Manipulation
  • Cooperative Manipulation of an Unknown Payload
  • Cooperative Attitude Control
  • Cooperative Control with Time-Varying Velocity
  • Aerial Manipulator Flight Test
• Learning for Control and Estimation
  • Drag Force Estimation
  • Contact Force on the Payload
  • Non-linear Adaptive Geometric Control

Safety-Critical Control — Current Research

Visiting Researcher, ETAIC Lab, UT Arlington · Tech Lead & Senior Controls Research Engineer, Amogy

Composing actuator-level safety architectures with state-space safety filters for end-to-end safety under embedded, real-time constraints.

• Safety certificates for low-level control loops.
• Safety-critical control for multi-agent systems.
• Safe control in the presence of model uncertainty and unmodeled dynamics.
• Safety-aware state-machine wrapper for legacy PID loops with provable input-boundedness, integrator-boundedness, and bumpless-transfer properties, evaluated head-to-head against control-barrier-function safety filters (in preparation, IEEE RA-L).

Autonomous Vehicle Planning and Control

Research Engineer, Ford Motor Company, Research & Advanced Engineering (advisor: Dr. H. Eric Tseng, NAE Member)

• Continuous-curvature (clothoid-based) path planner and nonlinear rear-wheel feedback lateral controller for autonomous parallel parking and auto-hitch (SAE L2/L3).
• State machines and control-barrier-function safety certificates for autonomous function transitions and fault handling.
• Trajectory-tracking benchmarks (pure-pursuit, LQR, PD, backstepping, nonlinear feedback); validated in simulation and on dSPACE real-time hardware.

Clothoid-Based Path Planning

Non-linear Rear-Wheel Feedback Control

Vehicle Dynamics with Cruise and Lateral Control

Pure-Pursuit Control

Autonomous Quadrotor UAV Control

Senior Controls Engineer, The Drone Racing League · Independent Research

• Cascaded PID, state-dependent LQR, nonlinear Lyapunov-based, sliding-mode, backstepping, geometric SE(3), and Cartesian impedance controllers for quadrotor trajectory tracking.
• Geometric attitude control and differential-flatness-based minimum-snap trajectory generation.
• End-to-end autopilot integrating state estimation, planning, and control, validated in MATLAB/Simulink, Gazebo, PX4 SITL, ROS, and on real quadrotor hardware.

Cascaded PID Control in PX4

State-Dependent LQR for Trajectory Tracking

Full-state time-varying LQR designed and implemented in real time in Gazebo and PX4.

Offboard Velocity Control

Minimum-Snap 3D Trajectory Generation

Distributed Control, Coordination, and Manipulation

M.S. research, Oklahoma State University (CoRAL Lab, Dr. He Bai), jointly with Dr. J. Á. Acosta (University of Seville, Spain)

• Decentralized multi-robot control framework for cooperative aerial manipulation: multiple quadrotor UAVs transport a shared payload without constant inter-robot communication.
• Adaptive force-sharing controllers regulate payload forces while all agents coordinate their motion, with stable transport under unknown payload mass and external disturbances — validated in simulation and physical flight tests.

Related publications:

• [J2] Thapa S., Bai H., Acosta J.A. Cooperative Aerial Manipulation with Decentralized Adaptive Force-Consensus Control. Journal of Intelligent & Robotic Systems (JINT), 2020.
• [C1] Thapa S., Bai H., Acosta J.A. Cooperative Aerial Load Transport with Attitude Stabilization. American Control Conference (ACC), 2018.
• [C2] Thapa S., Bai H., Acosta J.A. Force Control in Cooperative Aerial Manipulation. IEEE ICUAS, 2018.
• [C3] Thapa S., Bai H., Acosta J.A. Cooperative Aerial Load Transport with Force Control. IFAC NAASS, 2018.

Cooperative Manipulation of an Unknown Payload

Cooperative Attitude Control

Cooperative Control with Time-Varying Velocity

Aerial Manipulator Flight Test

Learning for Control and Estimation

M.S. research, Oklahoma State University (with Dr. Rushikesh Kamalapurkar)

• Concurrent-learning adaptive control for a team of robots transporting a common load, enabling real-time estimation of unknown parameters (payload mass and drag) while driving all agents and the payload to a desired trajectory.

• Guarantees parameter convergence and improves transient performance by reusing past data to relax excitation requirements, with accurate force regulation and synchronized motion in simulation.

• [J1] Thapa S., Self R., Bai H., Kamalapurkar R. Cooperative Manipulation of an Unknown Payload with Concurrent Mass and Drag Force Estimation. IEEE Control Systems Letters (L-CSS), with CDC presentation option, 2019.

Drag Force Estimation

Contact Force on the Payload

Non-linear Adaptive Geometric Control