Hardware system design of multi-sensor intelligent wheelchair

Abstract: In order to improve the ability of intelligent wheelchairs to obtain effective information in complex environments, a multi-sensor data acquisition system based on DSP is proposed. The system mainly includes: ultrasonic sensor, proximity switch, self-positioning sensor, posture sensor and visual sensor. This paper mainly analyzes and explains the system architecture and component design.
Keywords: intelligent wheelchair; DSP; multi-sensor system; TMS320LF2407 A

0 Introduction
The task of an intelligent wheelchair is to safely and conveniently take the user to the destination and complete the given task. During the movement, the wheelchair needs to accept the user's instructions and start its own obstacle avoidance, navigation and other functional modules in combination with environmental information. Unlike mobile robots, during use, the wheelchair and the user become a collaborative system. This requires that the human factor be taken into consideration at the beginning of the design. Therefore, safety, comfort and easy operation should be the most important factors in the design of intelligent wheelchairs. The difference in the physical ability of users determines that the intelligent wheelchair needs to be designed as an electronic system with diversified functions that can meet the needs of multiple levels. Modularity can best reflect the multifunctional characteristics of the system. Each user can choose the appropriate module integration according to the type and degree of his or her own disability, and the designer can easily improve the wheelchair function by adding functional modules on the existing basis. This article focuses on the modular design of intelligent wheelchairs.

1 Overall structural design of sensor system
The overall functions of the intelligent wheelchair can be divided into the following sub-functions: environmental perception and navigation function, control function, drive function and human-computer interaction function. Through the functional analysis and module division of the intelligent wheelchair, combined with the specific research content and expected control objectives, this system is mainly composed of three parts: sensor module, drive control module and human-computer interaction module. The hardware system structure is shown in Figure 1. The sensor module mainly consists of two parts: internal state perception and external environment perception. The posture information of the wheelchair itself is determined by the posture sensor; the self-positioning information is obtained by the displacement speed and distance of the encoder; the vision, ultrasonic wave and proximity switch are mainly responsible for continuously obtaining the distance information of the surrounding environment and obstacles. The drive control module adopts the rear-wheel drive mode. Each rear wheel is equipped with a motor to realize the forward, backward and steering of the electric wheelchair under the operation of the controller. The human-computer interaction interface is composed of two ways: joystick and personal computer interface data input to realize basic human-computer interaction functions.

Among them, the data acquisition unit plans to select DSP TMS320LF2407A as the control chip of the sensor module. TMS320LF2407A is a high-performance digital signal processor with high frequency and rich peripheral interfaces. Its main frequency can reach 150MHz, low power consumption (core voltage 1.8V, I/O voltage 3.3V); 128kX16-bit on-chip FLASH, 18kX16-bit on-chip SRAM, 4kX16-bit on-chip ROM; peripherals for motor control, 2 event managers; multiple standard serial port peripherals, 1 SPI synchronous serial port, 2 UART asynchronous serial ports, 1 enhanced CAN bus interface, 1 McBSP synchronous serial port; 16-channel 12-bit A/D converter; 56 independent programmable, multiplexed, general-purpose I/O ports. It can meet the requirements of this system design.

2 Multi-sensor data acquisition and processing
The intelligent wheelchair of this system has two independent driving wheels, each equipped with a motor encoder. The real-time detection data of the two motor encoders constitutes an odometer-type relative positioning sensor, and an inclination sensor and a gyroscope are installed to measure the posture state of the wheelchair during travel. Ultrasonic sensors and proximity switches are used to sense the surrounding environment information. In order to obtain obstacle information within a larger range, this system is equipped with 8 infrared sensors and 8 ultrasonic sensors. In addition, a CCD camera is installed to determine the depth information of the front travel path. The
hardware design schemes of the above sensors are introduced in turn below.
2.1 Ultrasonic sensor and proximity switch
This ultrasonic ranging system has a total of 8 ultrasonic sensors, which form an ultrasonic sensor array and are placed around the wheelchair. In order to detect some obstacles that are missed or not handled in time by the ultrasonic sensor, four inductive proximity switches are also installed around the wheelchair. When the obstacle hits the anti-collision rubber ring, the metal bar is deformed, resulting in vertical displacement, triggering the proximity switch action, and obtaining a switch signal (interrupt request signal), which immediately stops the mobile robot.
The ultrasonic environment detection circuit is mainly composed of a multi-channel analog switch, a boost amplifier circuit, a buffer amplifier and shaping circuit, and an ultrasonic transducer, as shown in Figure 2.

The boost amplifier circuit and the ultrasonic transmitting transducer constitute the ultrasonic transmitting part. The transmitting process is: first, the pulse width modulation channel of the DSP generates a modulated pulse wave with a certain pulse width, and after the transformer boost amplifier circuit, a momentary high-energy signal is generated to stimulate the ultrasonic transmitting transducer to generate an ultrasonic signal. It should be noted that at the moment of ultrasonic transmission, some sound waves will directly enter the ultrasonic receiving end, thereby generating a strong false reflection wave, causing the so-called ringing phenomenon. In order to avoid ringing, software delay processing is required, which leads to detection blind spots. In terms of program processing, the corresponding CAP interrupt is turned off for a period of time after the DSP transmits the excitation pulse wave, and the CAP interrupt is turned on after the blind spot interval has passed. The ultrasonic receiving part must work in coordination with the transmitting part to ensure accurate and sensitive signal reception. This part is mainly composed of ultrasonic receiving transducer, amplification filter, and shaping trigger output circuit. Since the energy of ultrasonic wave decreases with the increase of propagation distance during propagation, the echo signal reflected from a long-distance obstacle is generally weak, so it needs to be processed by multi-stage signal amplification before it can be detected by the DSP interrupt input port.
2.2 Encoder
In the intelligent wheelchair system, in addition to measuring the distance information of the environment, sometimes the orientation information must be effectively observed or estimated. For most indoor mobile robot systems, the orientation information is generally estimated indirectly through the information of the code disk, and this system also adopts this method. The result is obtained by calculating the information read from the code disk, but the cost is a certain amount of calculation time.
There are two time management modules (EV) on the TMS324LF2407A chip. Each EV module has an orthogonal encoding pulse circuit. After using this circuit, the orthogonal encoding pulse can be input on the two corresponding pins. This circuit can be used to connect the photoelectric code disk to obtain information such as the position and speed of the rotating machinery, but it should be noted that the capture function on the corresponding pin must be disabled at this time. The timing of
the orthogonal encoding pulse circuit can be provided by the general timer 2 (or general timer 4, EVB module). The general timer must be set to the directional increase/decrease mode and use the orthogonal encoding pulse circuit as the clock source.
The orthogonal encoding pulse is two pulses with changing frequencies and orthogonal (90° phase difference). It is generated by the photoelectric encoder on the motor shaft. The code disk is on the motor shaft and has many empty slots that can transmit light. When the motor drives the code disk to rotate, if the light emitted by the light-emitting diode is blocked, the photoelectric sensor behind it will not receive the signal, and then the photoelectric sensor will send a low-level pulse, that is, "0". If the rotation position is just right so that the light source can pass through the slots, the photoelectric sensor will sense the signal and send a high-level pulse, that is, "1".
The direction detection logic of the orthogonal encoding pulse circuit determines which of the two pulse sequences is the leading sequence, and then it generates a direction signal as the counting direction input of the general timer. Both edges of the two orthogonal input pulses are counted by the orthogonal pulse encoding circuit. Therefore, the generated clock frequency is 4 times that of each input sequence, and this clock is used as the input clock of the general timer 2 or 4. Figure 3 shows the waveforms of the orthogonal encoding pulse, the increase and decrease counting direction, and the clock.

2.3 Posture sensor
One of the most significant features that distinguishes this system from other wheelchair designs is that this design can balance the vehicle body with only two wheels. This significant feature requires it to have a special structure. The basic design idea is: keep the two wheels driven by independent DC motors and on one axis, keep the center of gravity of the vehicle body above the wheel axle, use the sensor that detects the inclination angle of the vehicle body to obtain the posture information of the vehicle body in real time, and the robot's processor processes the sensor signal, calculates the control amount according to a certain control algorithm to control the speed and direction of the motor, drives the robot forward or backward, and completes the balance of the vehicle body.
This intelligent wheelchair uses a combination of a tilt sensor and a gyroscope to form a posture sensor to detect the running posture of the vehicle platform. The tilt sensor is used to measure the angle of the wheelchair deviating from the vertical direction, and the gyroscope is used to measure the angular velocity. The motion controller
with TMS320LF2407 A as the control core calculates the control quantity through a certain control strategy according to the displacement and attitude signals of the platform operation detected by the encoder and attitude sensor, and then drives the DC motor to operate after pulse width modulation control and driver amplification, and adjusts the running speed of the vehicle platform at any time, so that the vehicle platform always maintains a balanced state. The control circuit schematic is shown in Figure 4. The control board collects signals from the inclination and angular velocity sensors and conditions the signals (filtering, shaping, offset), and then transmits the signals to the control board. After DSP calculation processing (the control algorithm is derived from the mathematical model of the electric vehicle system), the control signal is sent out through the two-way pulse width modulation of the DSP, and then the motor is driven by the motor drive module to operate and control the wheelchair to maintain a balanced state.

2.4 Camera
Used to perceive the depth information of the environment, such as determining whether there are stairs ahead and extracting the height information of the stairs, extracting road landmarks for navigation, etc. The camera can communicate directly with the PC via USB, which will not be described separately here.

3 Conclusion
This paper designs a multi-sensor environmental perception system for smart wheelchairs, and introduces each data acquisition subsystem in detail. It uses simple and reliable hardware circuits to perceive environmental information. Experiments have proved that this system solution has the characteristics of simple hardware circuit structure, reliable operation, high precision, good repeatability, etc., and it adopts a modular design, which can more conveniently add newly developed functional modules and carry out technical updates, so that consumers can choose and combine various modules according to their own life needs, so that each functional module can be fully applied, thereby meeting the needs of different consumer groups.