TL;DR
This paper presents a computer vision-based control system for an autonomous blimp designed to track moving or stationary targets. By combining a PD controller with the Camshift algorithm and Pulse Width Modulation (PWM), the system effectively overcomes the blimp's inherent nonlinearities and inertia, demonstrating robust tracking capabilities.
What Problem Does This Paper Address?
Controlling unmanned aerial vehicles (UAVs), especially those with significant inertia like blimps, presents a considerable challenge. Blimps are characterized by their large volume, low mass, and substantial inertia, leading to slow responses and complex dynamics. Furthermore, the ON/OFF nature of their DC motors introduces significant nonlinearities, making precise control difficult. Existing methods often struggle with these combined issues, particularly when dealing with real-time tracking of dynamic targets in an indoor environment where disturbances like air circulation can further complicate control. This paper aims to develop a robust control system that can accurately track targets despite these inherent limitations.
Methods
The core of the proposed system lies in a multi-component approach:
- Blimp Modeling: An approximate mathematical model of the blimp's dynamics is developed, focusing on the effective piece of its motion, which is dominated by inertia. Crucially, this model incorporates velocity and actuator saturation nonlinearities, which are critical for understanding the blimp's behavior.
- Computer Vision Tracking: The Camshift algorithm (a modified version of MeanShift) is employed for real-time object tracking. This algorithm is chosen for its ability to handle color-based histograms and track targets in video frames, even with some noise.
- Control Strategy: A Proportional-Derivative (PD) controller is designed to provide a stable control action. However, to address the ON/OFF nature of the DC motors and the resulting nonlinearities, Pulse Width Modulation (PWM) is integrated. The PWM acts as an interface between the linear controller output and the physical actuators, effectively approximating a continuous speed control.
- System Integration: A graphical user interface (GUI) is developed to allow for manual or automatic control of the blimp, facilitating the integration of the vision system, controller, and actuators.
Datasets
Not discussed in the paper. The experiments appear to be conducted in a controlled indoor environment using real-time video feeds from the blimp's camera.
Key Insights and Ideas
- Approximation is Sufficient: The paper demonstrates that a "decent approximation" of the blimp's behavior, particularly its saturation nonlinearities, is sufficient for effective control, rather than requiring a highly precise, complex model.
- PWM as a Nonlinearity Compensator: PWM is not just for speed control; it's presented as a crucial tool to bridge the gap between a linear controller (PD) and the inherently nonlinear ON/OFF actuators, enabling smoother control.
- Balancing Speed and Stability: The need to balance the blimp's speed for effective tracking with the requirement for slow, deliberate movements to avoid losing the target (due to camera FOV limitations and inertia) is a key consideration. This leads to the preference for a PD controller over a PID to minimize overshoot.
- Integrated System Design: The success hinges on the seamless integration of computer vision for target detection, a linear controller for stability, and PWM for actuator management.
Key Results
The paper presents experimental results demonstrating the system's ability to keep a tracked object within a defined region of the image frame. Figures 6, 7, and 8 show the output of the PWM commands for different directions (forward/backward, right/left) and the trajectory of the target in the image plane. The results indicate that the proposed controller, despite the blimp's inertia and disturbances (like air circulation), can effectively regulate the DC motors to maintain the target within the desired area.
Limitations and Future Work
- Limitations:
- The paper acknowledges that it's "essentially impractical to keep the tracked object at the exact center of the image" due to hardware limitations, motor time constants, and image processing delays. Performance specifications are relaxed to confine the target to a "small region."
- The ON/OFF nature of the DC motors and their minimum duty cycle requirement (80 ms) are significant constraints that the PWM aims to mitigate.
- The system's performance is sensitive to the wireless communication links between the camera, receiver, and motors.
- Future Work: The authors suggest expanding these methods to other platforms, such as quad-rotors.
Summary updated