In contemporary urban environments, the maintenance of high-rise building exteriors presents significant challenges, primarily due to the inherent risks and inefficiencies associated with traditional manual cleaning methods. As a researcher in aerial robotics, I have focused on developing innovative solutions to address these issues. The quadrotor drone emerges as a promising platform for high-altitude cleaning tasks, offering enhanced safety, efficiency, and cost-effectiveness. This article details my design of a quadrotor drone cleaning system utilizing the PixHawk open-source flight controller, which aims to revolutionize the high-altitude cleaning industry by replacing hazardous human labor with autonomous, reliable technology.
The high-altitude cleaning sector is relatively nascent yet fraught with dangers. Current practices predominantly involve suspended workers or specialized climbing robots, both of which suffer from limitations. Investigations reveal that many cleaning companies lack proper certifications, with approximately 70% of workers undergoing minimal training, leading to frequent accidents such as falls. Moreover, labor costs are escalating, and operations are often weather-dependent. Existing cleaning robots, including sliding sponge drones, rail-based systems, dual-rope-driven robots, and adsorption-type robots, struggle with low efficiency, restricted coverage, and high expenses. To overcome these hurdles, I have engineered a quadrotor drone-based cleaning system that prioritizes safety, performance, and affordability.
The core of my quadrotor drone cleaning system revolves around an ARM-Linux processor, which serves as the central processing unit. This processor executes four critical functions: First, it receives and decodes S.BUS protocol signals from the receiver, adjusting them based on real-time sensor data before transmitting the modified signals to the flight controller via an asynchronous serial port. Second, it continuously monitors system parameters, triggering alarms or risk mitigation measures when necessary. Third, it employs Pulse Width Modulation (PWM) to control four atomizing nozzles, one motor for the cleaning roller, and one water pump. Fourth, it stores sensor data in a local database and synchronizes it in real-time to an MQTT server for access by a ground control station. These capabilities ensure robust operation and adaptability in dynamic environments.
To facilitate reliable communication, I implemented a radio system centered on the S.BUS protocol. The S.BUS relay module acts as an intermediary, receiving signals from the remote controller, processing them with sensor inputs, and forwarding them to the flight controller. This allows for embedded secondary development of the flight controller, enabling real-time signal correction. The S.BUS protocol itself is a serial bus communication standard that transmits 16 proportional channels and 2 switch channels over a single signal line, using a non-standard baud rate of 100 kbps, 8 data bits, even parity, and 2 stop bits. Each frame consists of 25 bytes, with a start byte (typically 0x0F), an end byte (0x00), and 22 data bytes representing the channels. Each channel is encoded with 11 bits, providing a resolution of 2,048 steps, as shown in the formula for channel value extraction:
$$C_i = \sum_{j=0}^{10} b_{i,j} \cdot 2^j$$
where \(C_i\) is the value of channel \(i\), and \(b_{i,j}\) are the bits from the data bytes. This efficient protocol underpins the responsive control of the quadrotor drone.
For data management, I selected SQLite3 as the database system due to its serverless, lightweight nature and strong stability. It operates directly on files without requiring separate processes, making it ideal for embedded IoT devices like this quadrotor drone. The database logs all sensor readings, flight parameters, and cleaning operations, ensuring traceability and ease of maintenance. Table 1 summarizes the key features of the data storage module.
| Component | Description |
|---|---|
| Database Type | SQLite3 |
| Advantages | No server needed, zero configuration, lightweight, file-based operations |
| Compatibility | Embedded ARM-Linux systems |
| Data Types | Sensor data, flight logs, cleaning status |
The propulsion system of the quadrotor drone is crucial for sustaining flight with cleaning payloads. I employed the Hobbywing X6 power system kit, which includes electronic speed controllers (ESCs). With 23-inch propellers and a 12-cell battery setup, each motor can deliver a maximum thrust of 12.0 kg, resulting in a total thrust of 48.0 kg for the quadrotor drone. For optimal energy efficiency, I limit each motor to 5.0 kg of thrust, yielding a total of 20.0 kg, with a design redundancy that caps the maximum takeoff mass at 18.0 kg. The mass distribution of the quadrotor drone components is detailed in Table 2.
| Component | Mass (kg) |
|---|---|
| Frame (Yifeite G06Pro) | 5.0 |
| Power Batteries (2 sets of Gens Ace 16,000 mAh in series) | 1.8 |
| Water Tank with Water | 4.0 |
| Cleaning Device (4 atomizing nozzles, 1 roller) | 3.2 |
| Control System | 1.8 |
| Total Mass | 15.8 |
The power supply consists of two 16,000 mAh, 15C batteries with a standard voltage of 22.2 V connected in series. Based on field tests, this configuration supports approximately 50 minutes of high-altitude operation for the quadrotor drone, sufficient for extended cleaning sessions. The thrust-to-weight ratio is a critical parameter, calculated as:
$$TWR = \frac{T_{\text{total}}}{m \cdot g}$$
where \(T_{\text{total}}\) is the total thrust (20.0 kg for efficient operation), \(m\) is the total mass (15.8 kg), and \(g\) is gravitational acceleration (9.81 m/s²). For this quadrotor drone, the TWR is approximately 1.29, ensuring adequate maneuverability.
The mechanical design incorporates a custom-built cleaning apparatus attached to a robust frame. The quadrotor drone uses a Yifeite G06Pro four-rotor frame with 23-inch propellers, a 6 L water tank, a pump, a cleaning roller, and atomizing nozzles. The cleaning assembly is connected via acrylic clamps to carbon fiber rods, allowing quick detachment for transport. This modularity enhances the quadrotor drone’s practicality in diverse operational scenarios.
The cleaning control system is designed to adapt to varying contamination levels on surfaces like glass or curtain walls. The atomizing nozzles operate in three modes: Level 1 for lightly soiled areas with a wide spray angle, Level 2 for moderate contamination with increased water output, and Level 3 for heavy deposits like bird droppings, producing a focused water jet. These modes are controlled via PWM signals, with the roller speed adjusted accordingly through a dedicated receiver channel. During cleaning, the quadrotor drone can hover in place until the area is cleansed, then transition to stabilized flight for vertical movement.
The cleaning roller is the heart of the system, driven by two motors via flanges. A custom T-shaped rod with bearings connects the two rollers, preventing sagging and ensuring smooth rotation. This roller-based approach minimizes water absorption and reduces instability during flight, as it requires no additional buffering mechanisms against the cleaning surface. The atomizing nozzles, supplied by the pump, feature adjustable internal angles to optimize spray patterns. The pump’s performance is characterized by its power equation:
$$P = \frac{Q \times H \times \rho \times g}{\eta}$$
where \(P\) is power (W), \(Q\) is flow rate (m³/s), \(H\) is head (m), \(\rho\) is fluid density (kg/m³), \(g\) is gravity (9.81 m/s²), and \(\eta\) is pump efficiency. For my design, with a 12 V battery and 60 W max power, the pump achieves a head of about 1 m, adequate for effective cleaning. Table 3 outlines the nozzle modes and their parameters.
| Mode | Spray Angle | Water Output | Target Contamination |
|---|---|---|---|
| Level 1 | Wide | Low | Light dust, minor stains |
| Level 2 | Medium | Moderate | Moderate dirt, grime |
| Level 3 | Narrow (jet) | High | Heavy deposits, bird feces |
The water tank is a custom-designed component with a grid-like baffle system, resembling a “井” shape, that partitions the interior into nine interconnected compartments. This design mitigates liquid sloshing during flight maneuvers such as pitching, rolling, or flipping, thereby stabilizing the quadrotor drone’s center of gravity. The baffle allows water flow between compartments while reducing inertial effects. The tank includes a circular lid for secure mounting and a side outlet for pump connection. The mass of water influences flight dynamics, and the baffle helps maintain consistency, as described by the fluid dynamics equation for sloshing reduction:
$$\Delta F = m_{\text{water}} \cdot a_{\text{slosh}}$$
where \(\Delta F\) is the force variation due to sloshing, \(m_{\text{water}}\) is the water mass (4.0 kg), and \(a_{\text{slosh}}\) is the acceleration from liquid movement. The baffle minimizes \(a_{\text{slosh}}\), enhancing flight stability for the quadrotor drone.
To ensure reliable operation, I integrated a pump status detection mechanism. A brushed pump’s low-water signal is connected to the flight controller’s liquid level sensor port, outputting a high level when water is present and a low level when empty. Before takeoff, air is purged to prevent false alarms. Indicator lights signal various states: slow red flashing indicates no signal, fast red flashing denotes signal error, constant red means normal signal with pump off, constant yellow warns of low water, and constant green confirms adequate water supply. This system safeguards the quadrotor drone against dry running, which could damage the pump.
Autonomous cleaning via route planning is a key feature of this quadrotor drone. Leveraging the PixHawk 2.4.8 flight controller, I implemented waypoint-based navigation for efficient coverage. Using the QGroundControl (QGC) station, I define waypoints to create a “zigzag” pattern over building facades. The horizontal distance between adjacent waypoints at the same height equals the total length of two rollers, while vertical spacing matches the cleaning area’s height. This automates the quadrotor drone’s path, reducing pilot error and enhancing safety. The route planning algorithm involves GPS coordinates and altitude settings, with the flight controller managing autonomous takeoff, cleaning, and return-to-home functions. The mathematical representation for waypoint sequencing is:
$$W_i = (lat_i, lon_i, alt_i)$$
where \(W_i\) is the i-th waypoint with latitude, longitude, and altitude. The quadrotor drone follows these coordinates, with obstacle avoidance enabled by a guidance vision module. Table 4 summarizes the autonomous flight parameters.
| Parameter | Description | Typical Value |
|---|---|---|
| Waypoint Spacing (Horizontal) | Distance between adjacent points at same height | 2 × roller length |
| Waypoint Spacing (Vertical) | Height difference between levels | Cleaning area height |
| Return Altitude | Height for autonomous return | Set via QGC |
| Flight Mode | Autonomous grid pattern | Auto grid in QGC |
System testing involved both flight performance and cleaning efficacy evaluations. Initially, I conducted flight tests without propellers to calibrate the quadrotor drone, ensuring safe takeoff. With the water tank filled, the quadrotor drone demonstrated stable hovering and controlled maneuvers, confirming its airworthiness under load. The cleaning system was then activated, with the roller rotating smoothly and the nozzles spraying in all modes. The quadrotor drone successfully adhered to surfaces using a light-touch approach, minimizing pressure on walls. In autonomous mode, it followed predefined routes, simulating manual cleaning with consistent coverage. These tests validate the quadrotor drone’s capability to replace human labor in high-risk environments.
In conclusion, the quadrotor drone cleaning system based on PixHawk flight controller represents a significant advancement in high-altitude maintenance. By integrating robust propulsion, adaptive cleaning mechanisms, and autonomous navigation, this quadrotor drone addresses the safety, cost, and efficiency challenges plaguing the industry. Future work may focus on enhancing battery life, refining nozzle designs, and incorporating AI for real-time dirt detection. This quadrotor drone platform not only reduces casualties but also paves the way for smarter urban upkeep, demonstrating the transformative potential of aerial robotics in everyday applications.
The design process emphasized iterative optimization, with each component of the quadrotor drone tailored for reliability. For instance, the thrust calculations ensure that the quadrotor drone can carry its payload without strain, as expressed in the thrust equation for a single motor:
$$T = k \cdot \omega^2$$
where \(T\) is thrust, \(k\) is a motor constant, and \(\omega\) is angular velocity. For the quadrotor drone, balancing thrust across four motors is critical for stability, achieved through PID control in the flight controller. Moreover, the cleaning efficiency can be quantified by the coverage rate \(R_c\):
$$R_c = \frac{A_{\text{cleaned}}}{t}$$
where \(A_{\text{cleaned}}\) is the area cleaned and \(t\) is time. Compared to manual methods, this quadrotor drone achieves higher \(R_c\) due to continuous operation and precise control. Overall, the quadrotor drone system exemplifies how technology can mitigate human risk while improving productivity, offering a scalable solution for global high-rise maintenance needs.