Abstract: This paper, in accordance with the technical requirements of the first Freescale Cup National Undergraduate Intelligent Vehicle Invitational Competition, develops an unmanned vehicle navigation system based on dual-row, analog photoelectric sensors with forward-looking capabilities. It proposes a steering and drive control algorithm based on path memory and summarizes the experience in designing and manufacturing intelligent vehicles. Keywords: Unmanned vehicle; Navigation; Photoelectric sensing; Path memory Introduction In response to the Ministry of Education's call to strengthen the cultivation of undergraduates' innovative awareness, cooperative spirit, and innovative ability, the Department of Automotive Engineering at Tsinghua University actively formed a team to participate in the first Freescale Cup National Undergraduate Intelligent Vehicle Invitational Competition. Preparations began in December 2005, lasting eight months. Six generations of path recognition schemes based on photoelectric sensors were developed, an intelligent vehicle simulation research platform was built, and a steering and drive control strategy based on a path memory algorithm was proposed. Research was also conducted on power management, noise suppression, and drive optimization. Through extensive simulation experiments, road tests, and basic performance tests, an intelligent vehicle navigation system based on photoelectric sensing and path memory was developed, laying a solid foundation for the excellent performance of the entire vehicle system. This article will introduce the overall scheme of the intelligent vehicle, the selection of the path recognition scheme, steering and drive control, and the path memory algorithm. Intelligent Vehicle Overall Scheme The intelligent vehicle system is based on Freescale's MC68S912DP256 and consists of a power supply module, sensor module, DC motor drive module, steering motor control module, control parameter selection module, and microcontroller module, as shown in Figure 1. The intelligent vehicle system operates on a mixed voltage of +1.6V, +5V, and 7.2V. 7.2V powers the drive motor and steering servo, 5V powers the speed sensor, MCU, and photoelectric sensor receiver, and 1.6V powers the LEDs. For convenient online control parameter adjustment, a control parameter selection module is also provided. Different programs or control parameters can be called through several button settings to adapt to different site conditions. Figure 1: Overall Structure of the Intelligent Vehicle. The intelligent vehicle's operating mode is as follows: the photoelectric sensor detects track information, the speed sensor detects the current vehicle speed, and the battery voltage monitoring circuit detects the battery voltage. This information is then input into the microcontroller for processing. Control commands are issued to the race car through the control algorithm, and the movement trajectory and speed of the race car are controlled in real time through the steering servo and drive motor. In order to achieve good results in the intelligent car competition, the optimization of the model car chassis parameters and the reliability of the hardware equipment are very important. Among them, the optimization of the front wheel alignment parameters, the increase of the steering servo arm and the adjustment of the chassis center of gravity position have a great influence on the mechanical performance of the model car. The optimization of chassis parameters is described in [1], and will not be elaborated in this paper. Path recognition scheme selection and circuit design The path recognition scheme is the first thing to be determined, mainly including the following issues. * Photoelectric recognition or camera recognition; * How are the sensors arranged? What is the spacing, what is the shape, single row or double row; * The forward detection distance of the sensors; * Whether the sensor signal is digital or analog; * How to implement it in the circuit. Since the photoelectric recognition scheme is simple and reliable, the photoelectric recognition scheme is adopted in this paper. Digital Photoelectric Recognition vs. Analog Photoelectric Recognition The competition organizing committee requires a maximum of 16 sensors. Subtracting one speed sensor, 15 sensors are available for path detection. The allowed total width for sensor placement is 25cm. If digital photoelectric sensors are evenly distributed, the road detection accuracy can only reach about 17mm. This makes it difficult for the race car to achieve high control accuracy and response speed during movement. Essentially, the disadvantage of digital photoelectric sensors lies in their loss of a significant amount of path detection information. Analog photoelectric sensors, theoretically, can greatly improve path detection accuracy. The emission and reception of analog photoelectric sensors are both in a conical space with a fixed cone angle. The voltage magnitude is quantitatively related to the sensor's horizontal distance from the black path marking line: the closer to the black line, the lower the voltage; the farther from the black line, the higher the voltage (the specific correspondence depends on the phototube model and the height above the ground), as shown in Figure 2. Figure 2 Schematic diagram of the relationship between sensor voltage and offset distance Therefore, as long as the relationship between sensor voltage and offset distance is mastered, the distance between each sensor and the black mark line can be determined according to the magnitude of the sensor voltage (instead of just roughly judging whether the sensor is on the line), and then the position of the longitudinal axis of the vehicle body relative to the path mark line can be obtained, and the path information of continuous distribution can be obtained. According to the actual vehicle test, the accuracy of path detection can be improved to 1mm. In this way, the information collected by the sensor can ensure that the microcontroller can obtain accurate track information, thus providing a guarantee for improving the precise control of the race car. Double row arrangement and forward-looking design This paper developed an intelligent vehicle performance simulation platform [2] and conducted an in-depth study on the layout of sensors [3]. Since the steering servo, motor and car are all high-order inertial delay links, it takes a certain amount of time from input to output. The earlier the information of the road ahead is known, the more the lag from input to output can be reduced. Detecting the track a certain distance ahead of the car is called forward-looking. Within a certain forward-looking range, the larger the forward-looking sensor scheme, the higher its limit speed will be, and the higher its following accuracy of the guide line during high-speed driving will be, and the overall response performance of the system will be better. Therefore, the path recognition module is designed to be raised to form an angle with the ground. The front sensors are used for forward detection, while the rear sensors identify the starting point of the track and calculate the slope of the deviation between the vehicle's longitudinal axis and the track centerline to better adjust the vehicle's attitude. To ensure sufficient luminous intensity from the photoelectric sensors while maintaining maximum ground clearance, a high-current pulse-triggered luminescence control method is adopted. Experimental tests show that the current passing through the LED when it emits light is approximately 0.5A. If 15 sensors are used, the instantaneous current is 7.5A. Such a large current will definitely impact the battery voltage, which is detrimental to the normal operation of the entire system. Therefore, the luminescence times of the front and rear sensors are staggered, and luminescence is controlled by two sets of trigger circuits. This effectively reduces the impact on the battery voltage when the infrared LED emits light. Steering and Drive Control and Path Memory Algorithm Drive Motor Control This paper adds a geared disc to the motor output shaft, and the rotation of the motor output shaft drives the rotation of the geared disc. The photoelectric coupler and receiver are placed on both sides of the code disk. When the code disk rotates, the teeth on the code disk obstruct the light propagation emitted by the LED. Therefore, the resistance across the receiving tube will change significantly, resulting in a large change in the voltage across the sampling resistor in the circuit. By collecting the number of voltage pulses per unit time using the pulse capture port on the processor, the motor speed can be obtained, and thus the vehicle speed. The motor driver uses Freescale's MC33886. The difference is that this paper uses three MC33886s connected in parallel. This reduces the on-resistance, improves the motor's driving capability, and significantly improves the MC33886's heat dissipation; it also reduces the impact of the MC33886's internal overcurrent protection circuit on motor start-up and braking. The motor uses PID closed-loop control, which can adjust the PWM duty cycle in a timely manner according to different load conditions, allowing the vehicle to quickly track the target speed. To maximize vehicle speed, a maximum target speed is set on straightaways, with constant speed control. Speed ​​is reduced near curves, adjusted to the cornering limit speed when entering the curve, and accelerated in advance when exiting the curve. Steering Control Based on the current dual-row analog photoelectric sensor layout, the offset of the vehicle's longitudinal axis from the track centerline can be obtained, as well as the slope of the centerline relative to the vehicle's longitudinal axis. This allows for the determination of the vehicle's current attitude, which is then used for steering control. Here, the steering angle obtained from the front sensor signal is defined as θ1, and the steering angle obtained from the longitudinal axis slope information obtained from both front and rear sensor signals is defined as θ2. The final steering angle is determined by the formula: θ = k1θ1 + k2θ2. This control strategy enables weighted control of the vehicle's actual attitude, significantly improving cornering speed and reducing accumulated decision-making errors caused by detection accuracy issues. Furthermore, the combination of a large forward-looking sensor and a dual-row layout achieves early turning on normal curves and delayed steering on S-curves. To enable the servo to better respond to given steering angle values, PID control is employed. Parameter tuning is performed through road tests, ensuring high stability at high speeds. Path memory algorithm Since the competition rules require the vehicle to drive two laps on the track, the vehicle determines the straight road, curve, S-curve and the direction and radius of the turn by recording the number of pulses collected by the speed sensor and the steering servo in the first lap. Based on the data recorded in the first lap, the road points in the second lap can be segmented. The highest speed is used to accelerate on the straight road, and the vehicle is decelerated in advance before entering the curve to the maximum speed limit for the curve. Different speeds are selected for curves with different radii. The advantage of the path memory algorithm is that it can achieve the same effect as the CCD probe for complex S-curves, and use a small steering angle to pass through, which can greatly shorten the time. For the specific algorithm, please see [4]. Experience and conclusion The intelligent vehicle development work in this paper has gone through 6 rounds of development iterations, from the initial small look-ahead single row digital sensor to the pulse light emission, large look-ahead, double row arrangement, analog sensor scheme; the control strategy has been upgraded from simple PID control to path memory control, which has greatly improved the navigation performance of the vehicle. Some experience has been gained through the intelligent vehicle development process. * Initial development requires practical testing of the characteristics of photoelectric sensors, steering servo motors, drive motors, vehicle mechanical performance, steering sideslip characteristics, and battery characteristics. * Based on automotive theory, structural adjustments should be made to the vehicle within permissible limits to achieve optimal mechanical performance. * The organizing committee has developed a simulation platform; this tool should be fully utilized to study path recognition schemes based on photoelectric sensors. Combining hardware selection with experience in control and electronics, a path recognition scheme should be determined. Schemes with a longer look-ahead distance help improve vehicle throughput. * PID control is sufficient for vehicle requirements; parameter tuning needs to be combined with road testing. Acceleration and deceleration should not be too abrupt; smooth control yields good results. Excessive acceleration can cause overheating of the motor and drive chip, leading to decreased drive performance. This paper introduces the overall scheme, path recognition scheme selection, steering and drive control, and path memory algorithm of the winning car in the first National Undergraduate Intelligent Vehicle Competition. The use of a large-lookahead photoelectric sensor requires a significant current, resulting in substantial battery power consumption. Over longer track distances, the vehicle's battery level depletes rapidly, negatively impacting its racing performance. The fuzzy path-following algorithm based on path memory also requires improvement. Camera-based path recognition solutions, on the other hand, offer both large lookahead capabilities and lower power consumption, representing future directions for development. References: 1. Chen Song, Li Liguo, Huang Kaisheng, 'A Brief Analysis of Intelligent Model Car Chassis', Electronics World, 2006(11):150-153 2. Zhou Bin, Jiang Dinan, Huang Kaisheng, 'Intelligent Car Simulation System Based on Virtual Instrument Technology', Electronics World, 2006(3):132-134 3. Zhou Bin, Li Liguo, Huang Kaisheng, 'Research on the Influence of Intelligent Car Photoelectric Sensor Layout on Path Recognition', Electronics World, 2006(9):139-140 4. Zhou Bin, Liu Wang, Lin Xinfan, et al., 'Research on Intelligent Car Track Memory Algorithm', Electronics World, 2006(15):160-166 5. Huang Kaisheng, Jin Huamin, Jiang Dinan, 'Analysis of Korean Intelligent Model Car Technical Solution', Electronics World, 2006(5):150-152