One of the most important factors in surveillance systems using robots, is the complexity and unpredictability of the robot trajectories. This becomes more vital in hostile conditions where the robot trajectory is being followed by another agent. Therefore, random or chaotic sequences can be used in motion planning of surveillant robots. However chaotic sequences would be more effective due to their deterministic nature. Moreover the intrinsic robustness and ergodicity of chaotic systems, compared to random functions, would be another advantage to be considered in surveillance systems which require comprehensive coverage. In this paper, a method is proposed for chaotic motion planning for boundary surveillance and implemented to a quadrotor robot. Quadrotor robot is introduced as an appropriate choice for boundary surveillance application due to high maneuverability and aerial functions. The chaotic trajectory is produced using Henon map. Then the dynamics of the system is derived and a sliding mode controller is designed for such chaotic motion. Finally the dynamics of the robot and the proposed controller are simulated to generate the chaotic trajectories for two cases. The performance of the proposed algorithm is discussed according to unpredictability and staying in the allowable region. A circular path and a non-smooth path are considered for simulation examples.