1. Introduction

With the continuous development of technology, smart technology has become an indispensable part of modern automobile manufacturing. Intelligent technologies include automatic control, artificial intelligence, and big data. Applying these technologies can improve the efficiency, accuracy, reliability, and safety of automotive manufacturing, improving the quality and competitiveness of automotive products. Among them, intelligent technology based on vector controllers and wireless modules is one of the most advanced technologies in the current automotive manufacturing field. A vector controller is a controller used to control the motor speed and torque, which is used widely in automotive manufacturing ^[4]. With the constant development of Internet of Things technology, intelligent technologies based on vector controllers and wireless modules are used widely and researched. The wireless module is a communication technology used to connect various devices, with characteristics, such as low power consumption, high bandwidth, and reliability. It has also been used widely in automotive manufacturing ^[5,6]. Combining these two can achieve the application of intelligent technology in automotive manufacturing, improving the quality and competitiveness of automotive products ^[7]. A vector controller controls the motor speed and torque to control the vehicle driving speed, steering angle, and other parameters ^[8]. A wireless module is a communication technology that connects various devices (such as sensors and actuators) and transmits their signals to computers or other devices for processing and analysis. This research combines the improvement of vector controllers and intelligent technology to accelerate the implementation of unmanned workshops, ultimately achieving the front-end assembly completed by automatic robots. Therefore, this study focused on the vector control and optimization of electric tools as assembly improvements. In addition, this study integrated scene reconstruction and path optimization as the overall cruise control content of the workshop to explore the possibility of automatic operation in this mode. The research is significant in promoting the new application of smart technology in automotive manufacturing, improving the quality and competitiveness of automotive products.

2. Related Work

The vector controller is used widely in the assembly and manufacture of automobiles. The vector control device adjusts the torque size and angle in different motor devices to realize the dynamic control of the entire device. The exploration and improvement of this field have become a significant project in automobile manufacturing. Cheng et al. ^[9] reported that the controller can be used in different DC link voltages by proposing a current controller with vector proportional integration. The controller has certain advantages and effectiveness under different DC link voltages. Finally, the robustness and compensation capability of the coupled controller was demonstrated experimentally ^[9]. Chong et al. ^[10] evaluated the dynamics and natural servo control of multi-vector placement. The performance testing results showed that this method can meet tracking expectations and reject interference. Compared with traditional algorithms, the performance parameters are improved almost three times. Their research has solved the propulsion problem of multi-vector remote-control vehicles ^[10]. The experimental team of Goolak et al. investigated the operating conditions of asynchronous traction motor traction drives and performed performance simulations of vector control systems for current controllers and traction drives. The results showed the applicability of their improvements to system control ^[11]. Alipouri et al. ^[12] believed that the linear system has been studied and included the minimum variance benchmark method to evaluate the stability of the system. They proposed a support vector regression method to cluster analysis into the residual space of the process model, and the effectiveness and convergence of this method were proved ^[12]. Karimi et al. ^[13] improved the performance of the vector controller itself. They proposed a vector current controller with negative sequence current injection capability, which is simple and robust compared to the existing methods. The proposed controller could also achieve robust stabilization and active damping of the resonant poles ^[13]. Sun et al. ^[14] aimed to improve the efficiency of permanent magnet motors. They concluded that the traditional vector control method has decoupling errors, so they proposed a new approximate dynamic programming vector controller to solve this problem. The simulation and hardware experimental results showed that the proposed controller exhibited better performance in tracking reference and performed better for operating PM motors in linear and over-modulated regions ^[14].

Automotive manufacturing and assembly have changed, driven by the improved modern automation technology. Various hardware optimizations and upgrades of network systems have led to improved performance and the quality of the vehicle itself. Yadav et al. ^[15] used a melt deposition modeling process to design and analyze automotive parts. They reported that changes in the process parameters of the model would affect the mechanical properties. They also analyzed some effective parameters that varied at three levels, and their compressive and flexural strength would also be affected. This study is a meaningful exploration from the perspective of specific hardware ^[15]. Tian et al. ^[16] explored material manufacturing for automotive manufacturing and observed the parameters of each material under cold metal transfer wire arc additive manufacturing technology. Electron reverse diffraction showed that the material performance did not change significantly; its stability was better, and the substrate changed smoothly. The corrosion potential of the deposition layer was lower than that of the substrate, but the corrosion current density was higher. Their research catered to the material needs of automotive manufacturing ^[16]. Xu proposed a study on new energy vehicle repair in the context of economic, social, and technological development. He first elaborated on the value of the application of electronic diagnosis and focused on the considerations of this type of technology as a basis for electronics ^[17]. Yıldız’s team provided an optimization solution for the design of light vehicles, using a butterfly optimization algorithm to optimize the connection of rims with pins ^[18]. Qu Ying's team combined the development of modern data with automobile manufacturing. They suggested that intelligent manufacturing systems are the focus of the industry, but their understanding of SMS is insufficient. Therefore, they proposed a dynamic information service model to provide a reference for the industry. Their research generated ideas for the entire industry and contributed to the overall development ^[19].

Vector control is an important item in automotive manufacturing. Automatic and non-automatic torque control is throughout the whole process of automotive manufacturing. With the development of demand, its difficulty is also increasing. Adding intelligent technology to vector control can reduce manufacturing costs and make the entire process orderly. This study combines intelligent technology-based control systems with vector control to explore its feasibility in the manufacturing process.

3. Joint Construction of Vector Controller and Intelligent Technology

3.1 Optimization and Implementation of Vector Controller

Automotive manufacturing has moved into a human-machine integrated production model where automated assembly technology can save labor costs and increase accuracy. Steering control of intelligent assisted assembly robots is an issue to be considered because real-time adjustments in complex scenarios can increase the manufacturing floor compatibility. Electric tools are similar to autonomous vehicles. The interior, seat, and steering wheel must be assembled separately through direction control. The torque and tightening directions need to be adjusted in real time. The study will improve the power tools in the assembly chain from conventional motor power tools to remotely controlled vector control tools to meet the needs in terms of performance tuning ^[20]. The general purpose of brushless DC motors is to achieve an electronic phase change by checking the rotor position, which requires a high-position sensor to achieve the electronic phase change of the motor. Fig. 1 shows the vector control takes the required torque value as the optimal advantage to control electric tools and the control process.

Fig. 1. Standard DC motor vector control structure.

The conventional vector control in Fig. 1 usually uses three PWM outputs, which is difficult to implement because part of the arithmetic link of this control requires a certain amount of program initiation and is more demanding on the operator. Therefore, the study will use a microcontroller and hardware circuit instead of a programming link, improving the control efficiency. The vector control link generally involves three modes: $I_{d}=0$ control, maximum torque control, and weak magnetic field control. In $I_{d}=0$ control, the three-phase currents are measured by hardware; the currents are then transformed by coordinates, and the transformation equation is expressed as Eq. (1).

(1)

$ \left\{\begin{array}{l} I_{\alpha }=\frac{2\left(\cos 0.I_{U}+\cos 120.I_{V}+\cos 240.I_{W}\right)}{3}\\ I_{\beta }=\frac{2\left(\sin 0.I_{U}+\sin 120.I_{V}+\sin 240.I_{W}\right)}{3} \end{array}\right. $

After the right angle conversion of Eq. (1), the result was converted to $d-q$ using Eq. (2).

(2)

$ \left\{\begin{array}{l} I_{d}=\cos \theta .I_{\alpha }+\sin \theta .I_{\beta }\\ I_{q}=\cos \theta .I_{\beta }+\sin \theta .I_{\alpha } \end{array}\right. $

$\theta $ in Eq. (2) is the angle between the $d$ and $a$ axes. At the end of the conversion, the current and the rotor field will be in the same coordinate system, and the motor torque equation in this control method is expressed as Eq. (3).

(3)

$ T_{e}=p_{n}.\varphi _{fa}.i_{q} $

where $T_{e}$ is the electromagnetic torque. $p_{n}$ is the number of pole pairs. $\varphi _{fa}$ is the magnetic chain of the rotor. $f\mathrm{a}$ and $i_{q}$ are the $q$ axis current. In addition, the number of pole pairs and the magnetic chain of the rotor were fixed, so the torque is related directly to the $q$ axis current. Calculating the current will require a calculation of the motor power, which was transferred from the power supply, as observed with the structure. The electromagnetic power $p_{e}$ was calculated using Eq. (4).

(4)

$ p_{e}=u_{a}i_{a}+u_{b}i_{b}+u_{c}i_{c} $

where $u_{a}$, $u_{b}$, $u_{c}$, and $i_{a}$, $i_{b}$, $i_{c}$ are the components of potential and current on $a$, $b$ and $c$. respectively. The torque equation can be converted under the power calculation result combined with all kinds of equations, as shown in Eq. (5).

(5)

$ T_{e}=\frac{3\left(u_{d}i_{d}+u_{q}i_{q}+u_{0}i_{0}\right)}{2\theta _{rm}} $

$\theta _{rm}$ of Eq. (5) is the motor mechanical angle. $u_{d}$, $u_{q}$, $u_{0}$ and $i_{d}$, $i_{q}$, $i_{0}$ are the motor opposite potential component and current component on $d$ axis, $q$ axis, and 0 sequence, respectively. This control case is defined as $I_{d}=0$. Hence, the $d$ axis component is $0$, and the two components in the torque equation can be disregarded and simplified further as Eq. (6).

(6)

$ T_{e}=\frac{3u_{q}i_{q}}{2\theta _{rm}} $

Eq. (6) can reflect the relationship of each quantity. In the case of $\theta _{rm}$ constant, the sum of the torque $q$ axis of each component related to the component relationship is expressed as Eq. (7).

(7)

$ i_{q}=\frac{2\theta _{rm}T_{e}}{3u_{q}} $

where at constant speed and torque, the $q$ axis currents and counter potentials are related, and constant torque control can be achieved by establishing offline calculations and a database to apply commands to torque according to this principle. On the other hand, the details of this control, such as coordinate conversions, require a programming base. Hence, a microprocessor will be introduced to achieve unified control. This optimization method will not change the overall operation logic but will improve the control efficiency by leaving part of the process to the hardware ^[21]. The microcontroller used in the study integrates the FOC computing module with the chip and implements the calculations through the engine. The hardware composition of this controller has two motor drive circuits and two converters. The position angle signal of the controller was achieved using a Hall position sensor. When the motor moves, the sensor components sense different magnetic poles and transmit the generated level to the processing components. The element speed was determined by the number of pulses per unit time, which is a common way of sensing the engine speed. In addition, the motor drive was implemented by the internal drive circuit of the microcontroller, with a programmable PMD capable of motor control via voltage monitoring, achieving interaction with the converter. The circuit element has two modules, whereas the waveform generation circuit has functional circuits, such as pulse width modulation and conduction control. In addition, the working status of the motor needs to be monitored through the torque status. As mentioned above, torque control is achieved through component current, so comparing the standard current with it can achieve the purpose of monitoring. Therefore, modules are needed to determine the current. The circuit determination element used in this study consists of four independent voltage comparators. The working principle of this comparator was to compare the input signals and submit the comparison results to the microprocessor. At this time, the three phases of the microprocessor were used as input, and the output value after a series of arithmetic conversions was monitored.

The signal from the microprocessor will be used as the input set for the wisdom system, which will be transmitted through a wireless transceiver module and should have high integration and low external possession. In the whole framework, the main assembly realization link of the actual operation end of the operation will be done using this motor, and the wisdom module needs to complement the other functions outside the operation to ensure the normal environment of the assembly working conditions. The working principles of optimized vector controllers in automotive manufacturing are divided into four types: vector control, power-saving mode, speed control, and safety protection. Vector control refers to motor torque control, where the vector controller can convert torque from different directions into torque, enabling the motor to output the maximum torque. Power saving mode refers to the vector controller automatically selecting the optimal power saving mode based on the driving status and power demand of the vehicle to reduce battery consumption and pollution. Speed control refers to the vector controller that can adjust the speed and output power of the motor automatically based on the driving status and power requirements of the vehicle to achieve the optimal driving performance. Safety protection refers to the ability of the vector controller to automatically select safety protection measures based on the driving status and power requirements of the vehicle, such as overcurrent protection, undervoltage protection, and overvoltage protection. The optimized vector controller can control the torque of the motor to achieve energy-saving and improved power performance in the automotive manufacturing process.

3.2 IoT-based Intelligent Control Path to Achieve

The assembly of vehicles is designed as a single component, such as the vehicle interior, braking system, and steering system. In contrast, modern workshop assembly is usually carried out separately. Hence, intelligent technology is needed for workshop path planning. The study will use an automatic guided vehicle to guide the motor operation, and the path planning will be supported inline using intelligent algorithms. A demand analysis is needed before the path scheme. Taking the production line of a company as an example, the analysis showed that the automatic guidance should consider at least five production lines with nine sets of process equipment, so a 2D map modeling is needed regarding these production elements. Based on the needs of the workshop and the actual situation of guiding equipment in the transportation chain, the grid method is ultimately chosen to model the workshop map and the path planning will be based on the bionic ant colony algorithm ^[22]. Fig. 2 shows the mathematical model of the algorithm for the specific process.

Fig. 2. Basic process of the ant colony algorithm.

$P_{ij}^{k}$ supposes there are $m$ ants and they need to visit $n$ device points, the distance between device points $i$ and $j$ is denoted as $d_{ij}\left(i,j=1,2,\ldots ,n\right)\,,$ the information concentration between device points $i$ and $j$ at the moment of $t$ is denoted as $\tau _{ij}\left(t\right)$. $allow_{k}$ is the set of ants $k$ that need to visit the city. The probability that ant $k\left(k=1,2,\ldots ,m\right)$ moves from point $i$ to point $j$ at the moment $t$ is calculated using Eq. (8).

(8)

$ P_{ij}^{k}=\left\{\begin{array}{l} \frac{\tau _{ij}^{\alpha }.\eta _{ij}^{\beta }}{\sum _{s\in allow_{k}}\tau _{is}^{\alpha }.\eta _{is}^{\beta }},s\in allow_{k}\\ 0,otherwise \end{array}\right. $

$\eta _{ij}$ in Eq. (1) is the expected value of transfer from the point $i$ to the point $j$, which is generally $\eta _{ij}=1/d_{ij}$, and the importance is denoted as $\beta $. The value of $\beta $ is proportional to the dependence of the ant path decision on the heuristic function, when $\beta =0$, the path planning becomes a fully positive feedback algorithm. $\alpha $ is the degree of pheromone importance and takes a value proportional to the dependence of the ant decision on the pheromone concentration. When $\alpha =0$, the algorithm becomes a traditional greedy algorithm. $taub_{k}$ is the tabulation of the set of equipment points forbidden to be visited by the ants. The pheromone will be volatilized while visiting each device point, so the volatilization factor is added to optimize the pheromone problem in the path, and the pheromone accumulation problem is solved. The pheromone values were updated according to Eq. (9) when the ants visited all points.

(9)

$\left\{\begin{array}{l} \tau _{ij}\left(t+n\right)=\left(1-\rho \right)\tau _{ij}\left(t\right)+\Delta \tau _{ij}\left(t\right)\\ \Delta \tau _{ij}\left(t\right)=\sum _{k=1}^{m}\Delta \tau _{ij}^{k}\left(t\right) \end{array}\right.$

where $\rho $ is the pheromone play factor; $1-\rho $ is the pheromone residual factor; $t$ is time. $\Delta \tau _{ij}^{k}\left(t\right)$ indicates the pheromone increment of $k$ ants transferred from $i$ point to $j$ point in this cycle, which was calculated by three formulae: ant perimeter, ant density, and ant volume. According to the actual demand, this study used the ant perimeter model, and the calculation formula was Eq. (10).

(10)

$ \Delta \tau _{ij}^{k}\left(t\right)=\left\{\begin{array}{l} \frac{Q}{L_{k}},K ants to j in the cycle \\ 0,otherwise \end{array}\right. $

The constant $Q$ in Eq. (10) is the pheromone strength, which affects the algorithm efficiency, and $L_{k}$ indicates the length of the path taken by the ants $k$ in the loop. Under the ant colony algorithm, it is necessary to model the environment of empirical equipment points. The modeling abstracts the environment into a two-dimensional space. The model then divides it into 10${\times}$10 grids, which are divided into two categories: feasible and infeasible regions. The mapping relationship between the established map and coordinates is expressed as Eq. (11).

(11)

$ \left\{\begin{array}{l} x=r\left[\mathrm{mod}\left(i,N_{1}\right)-\frac{r}{2}\right]\\ y=r\left[N_{2}+\frac{r}{2}-ceil\left(\frac{i}{N_{2}}\right)\right] \end{array}\right. $

The $r$ of Eq. (11) is the area of the cell grid; $i$ is the grid number; $N_{1}$ and $N_{2}$ are the number of horizontal and vertical grids. After the grid was established, the path planning needed to be defined based on the requirements. In the path starting at $S$ and ending at $T$, path finding aims to find the available nodes $\pi \left(t\right)$ and satisfy $\left[0,t\right]\in X_{1}$, $\pi \left(0\right)=S$, and $\pi \left(t\right)=T$. Path $X_{1}$ represents the feasible path and does not conflict with the infeasible path $X_{2}$. The goal of path planning under this problem definition is to consider the length, path avoidance, and smoothness requirements, where the path length is calculated using Eq. (12).

(12)

$ \left\{\begin{array}{l} L=\sum _{i=1}^{n-1}\left\| c_{i+1}-c_{i}\right\| \\ \left\| c_{i+1}-c_{i}\right\| =\sqrt{\left(x_{i+1}-x_{i}\right)^{2}+\left(y_{i+1}-y_{i}\right)^{2}} \end{array}\right. $

$L$ of Eq. (12) represents the length; $n$ is the number of nodes in the path; $\left\| c_{i+1}-c_{i}\right\| $ is the Euclidean distance from the point $c_{i}$ to $c_{i+1}$. The smoothness of the path is a parameter that reflects the energy consumption of the robot under this path and has a direct relationship with the angle between the two paths. The smoothness of $smoothness$was calculated using Eq. (13).

(13)

$smoothness=\frac{1}{\alpha }\sum _{i=2}^{n-1}\left(\pi -\theta _{i}\right)$

where $\alpha $ is the number of turns; $\theta _{i}$ is the angle variable; the formula was calculated using Eq. (14).

(14)

$ \theta _{i}=\arccos \frac{\left(x_{i-1}-x_{i}\right)\left(x_{i+1}-x_{i}\right)+\left(y_{i-1}-y_{i}\right)\left(y_{i+1}-y_{i}\right)}{d\left(c_{i-1},c_{i}\right)d\left(c_{i},c_{i+1}\right)} $

where $x_{i-1}$, $x_{i}$, $x_{i+1}$, $y_{i-1}$, $y_{i}$, and $y_{i+1}$ are the horizontal and vertical coordinates of $i-1$, $i$ and $i+1$, respectively. $d\left(c_{i-1},c_{i}\right)$ and $d\left(c_{i},c_{i+1}\right)$ are the Euclidean distances of the three points. Another parameter considered is path safety $safety$, which is negatively correlated with the obstacles present in the path. The path is safer if the sum of the distance from the path to all obstacles is greater. During the specific implementation process, conflicts between the performance indicators may result in complex nonlinear results. Therefore, weights will be used to assign values to each parameter, and the fitness function $E\left(c\right)$ was calculated using Eq. (15).

(15)

$ E\left(c\right)=w_{1}.L+w_{2}.smoothness\;\;\; s.t \;\;\;safety\leq 0 $

$w_{1}$ and $w_{2}$ of Eq. (15) are the weights of path length and smoothness while $safety\leq 0$ suggests that the path is feasible and participates in the calculation as a constraint. After establishing the map model with the path planning definition, the ant colony algorithm will be used, and the entire intelligent control link will be implemented. According to the path algorithm and environmental requirements, the first node was calculated, and the search for the next node was continued according to the fitness function. At this point, the relationship between the number of ants and the initial number was determined. If it is less than, it indicates positive operation; otherwise, search again. The pheromone is updated when the phase is executed to determine if the number of iterations meets the requirements. The search will be stopped If it meets the requirements. Otherwise, the search will continue. When the path planning method automatically extracts the nodes on the way, the points are connected using curves to reduce the sharpness. This study used ant colony intelligence algorithms to plan production paths in complex environments, identify the optimal path, and improve the efficiency and stability of production lines, providing more efficient and high-quality production solutions for automotive manufacturing.

4. Analysis of Shop Floor Operation Performance of Vector Controller Joint Intelligence Technology

The workshop operation framework built in this study integrates vector control into the motor to realize the assembly of each component. The SCM reduces manual programming operations, realizes separation and interaction operations, installs the motor to realize the assembly environment, and realizes path planning based on the ant colony algorithm. The operation of the Internet of Things updates the environmental model in real-time using external sensors, providing defining conditions for path planning. Combining the two can achieve remote intelligent control and fully implement backend operations to avoid human interference with the environment. The performance analysis of the operation was divided into two parts: the assembly of the vehicle components, such as seats by the motor tool; the degree of arrival of the motor-equipped robot to the actual assembly scene. From the construction process of the vector controller, the torque is related to the component of the A-phase current, and the conventional vector controller is subjected to a MATLAB simulation of the current pulse and torque. Fig. 3 presents the results.

Fig. 3. Performance test of the traditional controller.

The simulation results of Fig. 3(a) show that the vector controller is more regular and can cycle, but there is a large perturbation in the torque output with a large up and down span. The current pulse situation in Fig. 3(b) was similar, but the burr of the waveform was larger, causing a non-constant torque, making it difficult to meet the requirements of fine assembly work. Hence, this paper proposes a simulation of the micro-control scheme according to the same environment. Fig. 4 shows the results of torque and current waveforms.

Fig. 4. Microcontrollers improve the performance of the appropriate controllers.

The microcontroller used in this study optimized the control of the motor. From the torque waveform shown in Fig. 4(a), the output torque was more stable under the same conditions, with little fluctuation and a small amplitude, which will help achieve specific assembly conditions. The current waveform output in Fig. 4(b) was also more stable and smooth, and the fluctuation within each waveform was not too large. Hence, the environment of this equipment can ensure stable output when the current vector is controlled. The study conducted output tests under a load torque of 1.0 N.m to determine the time required for the torque output to stabilize in a certain environment, as shown in Fig. 5.

Fig. 5. Time required for two controllers to stabilize the output.

According to the simulation results in Fig. 5, both controllers eventually achieved the requirements, but the time to reach the approved torque in Fig. 5(b) was less than in Fig. 5(a). Therefore, in terms of control time, the vector device optimized by the microcontroller reached the given requirement of torque output at 0.1s, and the conventional scheme took 0.4s. That is the output torque of the microcontroller reaches the set value faster than that of the conventional vector control. In summary, the microcontroller vector control scheme outperformed the conventional vector control in terms of both the stability performance of the current pulse and the time required to reach the approved requirements.

When the actual motor operation scheme is determined, the wisdom technology achieved by the motor will be measured. The evaluation mainly verified the effect of the ant colony path planning scheme constructed based on the map and provided a reference for actual operation.

As shown in Fig. 6, path planning involves three automatic robots reaching the assembly points of the frame, seat, and steering wheel, respectively, from the starting point. There is a possibility of overlapping paths 2 and 3, which is a critical point to overcome in the simulation. At this point, the assembly problem was planned based on the priority of the superior workshop. The workshop assembly prioritized the frame, steering wheel, and seats. Route 3 will be prioritized higher according to this requirement. Another result of the above simulation test was a certain smoothness in the handling of obstacle avoidance objects. In addition, the vehicles will not overlap in the conditions of taking the priority way to go out. At this point, another set of experiments will be conducted, as shown in Fig. 6.

Fig. 6. Path planning for three simultaneous paths.

The result of Fig. 7 in paths 2 and 3 is the case. Under the priority assignment, the automatic vehicle of path 2 will be reached in priority if it is difficult to avoid a collision between two paths under the working requirements in the phase direction. Therefore, in the case of high-priority routing, it is necessary to redraw the path planning for low-priority, and ant colony algorithm-based routing can effectively avoid rerouting. Table 1 lists the specific parameters of the above experimental path planning design.

Fig. 7. Planning results under cross-paths.

From the information of six path planning in two sets of experiments in Table 1, obstacle avoidance path planning under a priority-constraint strategy can avoid finding more complex detour solutions while minimizing the total length of the path, saving workshop processing resources. Path planning will be performed with three kinds of obstacles set up to test the obstacle avoidance function of the method in the workshop, as shown in Fig. 8.

Fig. 8. Obstacle avoidance test of path planning scheme.

From the obstacle avoidance path planning in Fig. 8, this path planning has good obstacle avoidance performance in practice when there are dynamic disturbances, such as worker and fixed component interference. This softens the steering sharpness while avoiding three obstacles within the feasible area, reducing driving energy consumption. In practice, the IoT-based sensor control will monitor the obstacle situation in real time, and updating the path planning, in this case, can achieve new path turnover. Performance testing was conducted to explore the planning speed and accuracy of this method while ensuring its effective path-planning strategy. This study used artificial path planning for comparison, as shown in Fig. 9.

Fig. 9. Performance analysis of path planning scheme.

The overall error of the ant colony algorithm path search scheme was 0.1-0.13 (Fig. 9), and the fluctuation range was not large, which was related to the obstacle setting. The error fluctuated in a specific range of approximately 200s. Compared with manual search, the overall error level was higher than that of the ant colony algorithm, fluctuating between 0.1 and 0.15, with an extensive fluctuation range and unstable performance. Planning stagnation is prone to occur if there are dynamic obstacles. The application effect of the proposed improved swarm algorithm in dynamic environments was examined. Performance comparison experiments were conducted using similar sparrow search algorithms, flying mouse search algorithms, and traditional ant colony algorithms. This experiment used the optimal path length, search success rate, and search duration as performance comparison indicators. Table 2 lists the performance comparison results of the three algorithms.

Table 2 lists the performance comparison results of the three algorithms. In Dynamic Environment 1, the optimal path length of the improved ant colony algorithm was 24.26m, which was lower than the 25.38m, 25.27m, and 26.68m for the sparrow search algorithm, flying mouse search algorithm, and the traditional ant colony algorithm, respectively. Moreover, its search time was 6.94s, which was less than that the 10.38s, 11.25s, and 15.86s of the sparrow search algorithm, the flying mouse search algorithm, and the traditional ant colony algorithm, respectively. In dynamic environment 2, the optimal path length of the improved ant colony algorithm was 23.38m, which was lower than the 24.26m, 24.23m, and 25.45m of the sparrow search algorithm, the flying mouse search algorithm, and traditional ant colony algorithm, respectively. The search success rate was 99%, higher than the 98%, 98%, and 96% of the sparrow search algorithm, the flying mouse search algorithm, and the traditional ant colony algorithm, respectively. In Dynamic Environment 3, the optimal path length of the improved ant colony algorithm was 24.49m, which was lower than the 25.57m, 25.38m, and 26.83m of the sparrow search algorithm, the flying mouse search algorithm, and the traditional ant colony algorithm, respectively. The search time was 6.85s, which was lower than the 10.66s, 11.68s, and 16.35s of the sparrow search algorithm, the flying mouse search algorithm, and the traditional ant colony algorithm, respectively. These above results suggest that in the three dynamic environments, the optimization performance of the improved ant colony algorithm was better than that of the comparative algorithm. Therefore, applying it to workshop path optimization can improve the workshop work efficiency. Finally, a specific processing and production workshop was studied, and the combination of the motor and automatic path robot was tested. The operating environment and parameters were set and put into use. Assembly experiments were conducted on the steering wheel, seat, and frame, and the assembly was successful, as shown in Fig. 10.

Fig. 10. Combination of Vector Motor and Automatic Robot for Automobile Assembly.

Fig. 10 presents the experimental test of the assembly realized by the automatic path equipped with a vector motor, which is the actual test of putting the above two automatic operations into use. The assembly scores of all three components were above 60 points. The highest score appeared in the assembly experiment of the frame and seat. Hence, the intelligent technology of the joint vector controller constructed in this study can at least meet the installation requirements of the frame and seat.

5. Conclusion

Most modern automotive manufacturing has taken semi-mechanized operations to replace the manual assembly. This automation trend has room for further improvement because of technological advances. The research takes intelligent control from two permutations to realize a fully automated unmanned workshop. One was the specific assembly to realize the process of vector motor automation, which replaces manual programming with embedded micro-control while improving performance. The second was to use an automatic robot to transport the vector motor to the operating platform and use the ant colony algorithm for path planning in the workshop where the working conditions are determined. Real-time path planning was achieved through an update of the sensor-based IoT platform. In the experiment of result analysis, the performance of the vector motor was first analyzed. These results suggest that the vector controller under the microprocessor can generate a more stable torque output at 0.1 seconds. In the obstacle avoidance experiment of automatic transportation robots, path planning can consider the workshop component and dynamic obstacles for path planning. The planned path corner sharpness is relatively low. The error analysis results suggest that the error of this method fluctuates between 0.1 and 0.13. The hands-on assembly with a transport robot equipped with vector motors shows that the method can satisfy the assembly of seats and frames. The next step in the research will be to optimize the combination of transportation robots and vector motors, stabilize the performance through physical means, and achieve more precise operations.