This project presents the design and validation of a reactive control architecture for a quadrotor UAV performing person-following in cluttered environments.
The UAV does not navigate toward a static goal. Instead, it maintains a desired relative pose with respect to a moving target (Bill), following it at a fixed distance and altitude offset. Obstacle avoidance is handled independently from tracking through a combination of repulsive fields and vortex (tangential) fields, without introducing any classical attractive potential toward a goal.
The system is fully reactive and does not rely on global path planning for the UAV. The drone continuously:
• Computes a virtual setpoint behind the target • Maintains a desired following distance • Reacts to obstacles detected through onboard proximity sensors
Obstacle avoidance is based on two complementary components:
Pushes the UAV away from obstacles when entering their influence region. This guarantees collision avoidance but alone may cause oscillations or local minima.
Adds a rotational component around obstacles. Instead of pushing the drone straight back, it generates a lateral sliding motion that guides the UAV smoothly around obstacles.
The final obstacle-induced action is:
F_obs = F_rep + F_vor
This ensures both safety (repulsion) and smooth circumnavigation (vortex).
To avoid instability and aggressive corrections, several heuristic stabilizations are implemented:
• Smooth blending between tracking and avoidance • Cubic shaping of avoidance activation • Low-pass filtering of avoidance forces • Saturation of lateral forces • Hysteresis logic to prevent chattering
These mechanisms ensure that obstacle avoidance acts as a bounded perturbation, rather than dominating the tracking objective.
The repository includes four different simulation scenarios, progressively increasing in complexity:
Baseline person-following behavior. The UAV maintains the desired relative pose without interference.
Repulsive + vortex fields active but without smoothing/blending. The system shows instability, sensitivity to perturbations, and non-repeatable trajectories.
Full avoidance strategy with smoothing and blending enabled. The UAV smoothly bypasses obstacles and consistently re-aligns behind the target.
Narrow passage with continuous obstacle interaction. The UAV performs micro-corrections, maintains bounded tracking error, and safely exits the corridor.
The repository includes:
• Complete technical report • Simulation files • Parameter configurations for all scenarios • Presentation material
This project demonstrates that combining repulsive and vortex fields, together with proper stabilization heuristics, enables robust real-time UAV person-following in complex environments without requiring global path planning.