DC motor as the power source has fine drive ratio that makes speed adjustment easier to find a perfect speed to coordinate with the camera video acquisition and WiFi transmission speeds [4E.A. Ramadan, M. El-Bardini, and M.A. Fkirin, "Design and FPGA-implementation of an improved adaptive fuzzy logic controller for dc motor speed control", Ain Shams Eng. J., vol. 5, no. 3, pp. 803-816, 2014.
[http://dx.doi.org/10.1016/j.asej.2014.04.002] ]. It is in small size with compact structure and strong ability to withstand overloads, which satisfies the requirements for a robot to work in a narrow and harsh environment. The equipment is in new type with good protection performance for continuous working in corrosive, damp and other harsh environment.

By using differential crawler movement manner [5B. Maclaurin, "A skid steering model using the magic formula", J. Terramechs., vol. 48, no. 4, pp. 247-263, 2011.
[http://dx.doi.org/10.1016/j.jterra.2011.04.002] ], the left and right tracks are respectively controlled by wheels driven by two motors, and each wheel is connected to the robot body via a spring damping device. Since the space under a vehicle is very limited, in order to efficiently detect the vehicle chassis, the robot must be able to perform flexible steering with small turning radius. Differential crawler movement manner is advantageous in flexible movement and convenient control, which can help complete situ rotation and other difficult manoeuvre. It can adapt to various workplaces and maintain the highest stability in a complex space, providing the basis for collecting stable chassis videos at the mean time of running. When the robot runs too fast, the image quality might be poor due to performance of the camera and WiFi transmission efficiency, leading to poor detection results. While using the PWM control motor, there might be a driving force shortage when the motor speed is below a certain threshold, leading to poor obstacle climbing performance. Therefore, a mechanical gear set (Fig. 3) has been used to slow down the motor.

Fig. (3) The gear set of the robot mobile system.

According to the design requirements, the gear set transmission ratio should be between 1/7 to 1/5. Limited by the structure of the robot, a set of gears as shown in Fig. (3) have been designed, where the gears 4/5/6/7 satisfy the features of concentric gear wheel system:

(1)

Experiments showed that the robot performance is better when the speed gear ratio is between 5 and 7, namely:

(2)

(3)

(4)

(5)

The motor rotational speed is stabilized at 500rpm using PWM control, and then adjusted to 85rpm using this speed gear set. Practice has proven that the robot with this gear set is stable and robust and is capable to overcome the problem of insufficient power at low speeds.

Two servo motors controlled by SCM are used to drive the camera platform, powered by a servo module that drives the two independent rotating surfaces arranged perpendicular to each other. Scroll bars of the platform control variation of the two servo angles of the two independent perpendicular rotation planes, so that one motor drives camera in a horizontal plane and the other drives in the vertical plane. The camera is mounted on the top of the platform and the orientation can be freely adjusted, providing an unrestricted field of vision to complete a full range of chassis detection.

Fig. (4) Portion of the motor drive control circuit.

The motor is driven by an H-bridge circuit that was built upon two BTN7971 chips in a simple design, requiring only one PWM control channel [8M.B. Ahrens, J.M. Li, M.B. Orger, D.N. Robson, A.F. Schier, F. Engert, and R. Portugues, "Brain-wide neuronal dynamics during motor adaptation in zebrafish", Nature, vol. 485, no. 7399, pp. 471-477, 2012.
[PMID: 22622571] ], 2012). The on-chip power consumption is relatively small, leading to low heat and stable performance, so that the affection by the power heat can be effectively avoided. The direction of motor current can be controlled, providing advantageous performance for emergency stop and quick start. The motor output is given by

(6)

In the above equation, n is the rotational speed of the DC motor, U is the armature voltage, I is the armature current, R is the total resistance of the armature, φ is the magnetic flux, K is structural parameters of the motor. This formula indicates that motor speed can be controlled by the excitation magnetic flux or armature voltage. The equivalent armature voltage control method is adopted based on PWM waves for DC motor speed adjustment.

(7)

The PWM control intends to change the duty cycle of a square wave signal, i.e. the conduction time t₁ during a cycle, when the supply voltage Us is constant. In this case, Uo is the equivalent armature winding terminal voltage of the motor, D is the duty cycle representing the conduction time period ratio in a cycle (0≤D≤1). Portion of the driving circuit is shown in Fig. (4).

Automobile chassis inspection robot requires stable power supply, for which reason three-terminal regulator chips are adopted to construct a simple circuit (Fig. 5). Capacitors were used between input/output voltages of the chips and the ground, which filters out the current fluctuations with workloads and improves the power stability.

Fig. (5) Stabilized power supply circuit.

The WiFi communication relies on a TP-LINK WR703N wireless router, which has been updated with OpenWrt firmware to make the router work like a small Linux system, adaptive to privately developed driver software [9M. Jingli, and L. Teng, "Design of smart home server based on openwrt", Network Security Technol. and Appl., vol. 11, pp. 197-198, 2014.]. Through modification, TX / RX serial communication was added, using which the router system and drive software can be re-written. After the corresponding camera driver was installed, video images can be captured via a USB camera. Meanwhile the communication protocol was updated after modification via serial ports [10V.S. Malladi, A. Mishra, and Z. Ji, "System and method of determining a location based on location of detected signals", US, US 8150367 B1, 2012.], so that two-way real-time transmission of video data and control data can be completed simultaneously.

The control platform is developed into exe software running on the remote computer, which consists of menu bar, video display area, control buttons and slid bars as shown in Fig. (7):

The platform client terminal exchanges data with the robot routing module via TCP/IP communication protocol. The control platform can call the robot camera to obtain images in real time, with saving and playback functions [11N.T. Luy, N.T. Thanh, and H.M. Tri, "Reinforcement learning-based intelligent tracking control for wheeled mobile robot", Trans. Inst. Meas. Contr., vol. 36, no. 7, pp. 868-877, 2014.
[http://dx.doi.org/10.1177/0142331213509828] ]. The communication protocol is stored in the robot system control platform. After successful logging into the control platform via WiFi router, a new thread immediately starts, waiting for control platform to issue control instructions. The microcontroller system responds after the instructions are received.

Fig. (7) Control platform diagram.

The robot workflow is as shown in Fig. (6).

1. The robot powers on, the WiFi routing communication module starts broadcasting WiFi signals, and waits for the robot control platform to build connection and control.
2. Run the control platform, search for and connect to the robot via WiFi.
3. The control platform confirms WiFi connection is completed, and prompts to select the control mode. Initialization of each module is completed, and the vehicle chassis inspection robot finishes preparation works.
4. Control the robot to move back and forth or left to right via the WSAD four-button keypad or arrow keys on the control platform. It is then possible to control the orientation of the camera platform and the robot hand to perform corresponding actions by dragging the slide bars in the control platform.
5. The vehicle chassis inspection robot can automatically run along with designed routes and collect video images at a certain site, and transmit the videos to the control platform via WiFi for storage. It is also possible to artificially control the robot, perform visual check of vehicle chassis at a specific position. The inspection status is as shown in Fig. (8).

Fig. (8) The inspection status of the robot.

When the robot has just completed structures, many disadvantages in structural design started to emerge, leading to unstable operation. The camera coordinates with the entire system in an imperfect manner, often jams, while the videos are sometimes discontinuous. It was found that the root problem is that the transmission structure is somehow irrational and the movement is unstable. The robot has then been revised according to the test results in varying environment.

During experiments, the robot was driven with categorized speed ranges for different road conditions, including a flat concrete floor, a road with different degrees of potholes, roads with different slopes. It was found that the motor speed limit when running on a flat road surface is 115rpm and the camera collected images exhibit jam phenomenon when the speed is above this limit. For the road with potholes, the video jitters seriously when the speed is more than 98rpm, which cannot satisfy the chassis inspection purposes. On the uphill sections, the drive motor stops working when the speed is below 48rpm, and the optimum speed is between 58rpm and 125rpm. According to these experimental results and comprehensive analysis, a gear set has been designed to provide better speed performance.

Some video screenshots captured by the robot are shown in Fig. (9).

Fig. (9) Image captured by the chassis inspection robot.

As shown above, the vehicle chassis inspection robot can provide results that satisfy the requirements of the automotive service industry, fully meet the design objectives. During automatic inspection, the influence of light intensity on the robot detection results is given below Table 1

Table 1
Influence of different light intensity.

Experimental results showed that, between 8:30am and 16:30pm, 96% area of the car chassis can be clearly observed by the robot, which means that the robot can be operated in full order to provide good detection results.

The Table 2 shows how different routes affect time spent on inspection.

Table 2
Time expenditure affected by routes.

When S route as shown in Fig. (10) is used for chassis inspection, the vehicle chassis including inner surface of the tires can be fully detected without dead vision space, leading to good results and high efficiency. After comparison with the data collected from vehicle service companies, it was found that 40% inspection time can be saved after using this robot.

Fig. (10) Robot working site and route.