Zero-Shot, Safe and Time-Efficient UAV Navigation via Potential-Based Reward Shaping, Control Lyapunov and Barrier Functions

2026-05-03

AI Summary

This work addresses the challenges of prolonged task duration and insufficient safety in autonomous drone navigation within complex environments by proposing a reinforcement learning approach that integrates potential-based reward shaping (PBRS) with control Lyapunov/barrier functions (CLF/CBF). The method trains a policy in a simple environment and leverages a CLF-CBF-QP filter to enable zero-shot transfer to complex scenarios, ensuring both safety and task efficiency without requiring retraining. By unifying PBRS with the CLF/CBF framework—a novel combination to date—the approach significantly reduces mission completion time in simulations while demonstrating superior safety and navigation performance in intricate environments.

Abstract

Autonomous navigation and obstacle avoidance remain a core challenge of modern Unmanned Aerial Vehicles (UAVs). While traditional control methods struggle with the complexity and variability of the environment, reinforcement learning (RL) enables UAVs to learn adaptive behaviors through interaction with the environment. Existing research with RL prioritizes the mission success at the expense of mission time and safety of UAVs. This study integrates Potential Based Reward Shaping (PBRS) with Control Lyapunov Functions (CLF) and Control Barrier Functions (CBF) to simultaneously optimize mission time and ensure formal safety guarantees. An RL model is trained in a generalized simple environment, then used in complex scenarios incorporating a CLF-CBF-QP filter without further training. Experimental results in simulated environments demonstrate a significant reduction in mission time and outstanding performance in complex environment.

Problem

• UAV navigation
• obstacle avoidance
• safety
• mission time
• reinforcement learning

Innovation

• Potential-Based Reward Shaping
• Control Lyapunov Function
• Control Barrier Function
• Zero-Shot Navigation
• Safe Reinforcement Learning