Authors: Fu Yili, Cao Zhengcai, Wang Shuguo, Jin Bao

I. Introduction

In order to work autonomously in an unknown or time-varying environment, a multi-joint robot should have the ability to feel the working environment and plan its own actions. Therefore, it is necessary to improve the robot’s ability to quickly understand and recognize the current perception environment and avoid obstacles in real time. Real-time obstacle avoidance is the key technology to realize the autonomous working ability of intelligent robots. It is also a hot spot in the recent development of intelligent robots at home and abroad. Its distinctive feature is that it has sensor information feedback and can achieve good intelligent behavior. This paper mainly focuses on the research on the real-time obstacle avoidance method of multi-joint robot based on sensor information. The selection of sensors and sensor information fusion technology are introduced in detail.

2. Sensor selection

One of the key issues in robot obstacle avoidance is how to use the sensor’s perception of the environment during motion. Any type of sensor has its own advantages and disadvantages. Various factors need to be carefully considered when selecting.

In the process of robot motion planning, sensors mainly provide the system with two kinds of information:

(1) The existence information of obstacles near the robot.

(2) The distance between the obstacle and the robot. In recent years, sensors applied to robot motion planning are generally divided into two categories: passive sensors and active sensors.

1. Passive sensor

Passive sensors used in obstacle avoidance include tactile sensors and visual sensors.

(1) Tactile sensor

The robot haptic system is a sensory function that simulates the contact between human skin and objects. It obtains information about the surrounding environment. It is used to avoid obstacles. Especially in the dark or under the condition that the information cannot be obtained visually due to the influence of obstacles. Has haptic functionality.

A tactile sensor is a device that measures the parameters of its own sensitive surface interacting with external objects. A tactile sensor often contains many tactile sensitive elements and is arranged in the form of an array. The tactile image is generated by mutual contact between these tactile sensitive elements and objects. Analysis and processing. This way of working is called passive haptics/but. In practical applications. On the one hand, the spatial resolution of tactile sensors is greatly improved.

The size of its working plane is much smaller than that of the recognized object; on the other hand, the three-dimensional information of the object needs to be obtained in the robot control. Therefore, on the basis of passive haptics, the tactile sensor is installed on the robot. With the continuous movement of the robot, the sensor can obtain the three-dimensional tactile information of the recognized object. After further processing and recognition, it is reflected to the robot controller. It enables the robot to obtain the information of the surrounding environment, identify the shape of the object, determine the spatial position of the object, etc., so as to achieve the purpose of intelligent control and obstacle avoidance. This way of working is called active haptics. When installing tactile sensors, they are generally installed on major operating parts such as claws, feet, and joints.

The main defects of tactile sensor application in the multi-joint robot obstacle avoidance system are: signal lag, it is difficult to achieve real-time obstacle avoidance, and the robot system is easily damaged during the work process.

(2) Vision sensor

The amount of information obtained by vision sensors is much more than that obtained by other sensors, but it is far from being able to make robot vision have the same functions as human beings. Generally, the development of vision sensors is limited to the functions required to complete special operations.

The visual sensor converts the optical image into an electrical signal, that is, converts the spatially distributed light intensity information incident on the photosensitive surface of the sensor into an electrical signal that is serially output in time series – a video signal, and the video signal can reproduce the incident optical radiation. image. There are three main types of solid-state vision sensors: one is a charge-coupled device (CCD); the second is a MOS image sensor, also known as a self-scanning photodiode array (SSPA); the third is a charge injection device (CID). At present, the CCD camera is widely used in the robot obstacle avoidance system, which can be divided into two types: linear array and area array. The linear array CCD captures one-dimensional images, while the area array CCD can capture two-dimensional plane images.

The image captured by the vision sensor is converted into a grayscale matrix after spatial sampling and analog-to-digital conversion, and is sent to the computer memory to form a digital image. In order to obtain the desired information from the image, it is necessary to use the computer image processing system to perform various processing on the digital image, and send the obtained control signal to each actuator, so as to reproduce the control of the multi-joint robot’s obstacle avoidance process.

This sensor has three main defects in obstacle avoidance: first, it is limited by light conditions and working range; second, the driving circuit of this type of sensor is complex and expensive; third, the real-time performance is poor.

2. Active sensor

Active sensors are divided into ultrasonic sensors, capacitive coupling sensors, eddy current sensors, and infrared sensors due to different intermediate transmission media.

(1) Ultrasonic sensor

Ultrasonic sensors rely on emitting a certain frequency of sound wave signals, and use the ultrasonic reflection on the interface of the object to scatter to detect the existence of the object. If ultrasonic waves encounter other media when propagating in the air, they will be reflected due to the difference in acoustic impedance of the two media. Therefore, ultrasonic waves are emitted to the measured object in the air, and the reflected waves are detected and analyzed to obtain information on obstacles.

Ultrasonic sensors are widely used in tasks such as robot ranging, localization, and environment modeling due to their simple, fast, and low-cost information processing. However, there are certain limitations in the real-time obstacle avoidance system of multi-joint robots, mainly in four aspects:

First, because the wavelength of ultrasonic waves is relatively long, specular reflection can occur for slightly larger flat obstacles, and the sensor cannot receive the reflected signal, so the obstacle cannot be detected.

Second, the blind area is large, because each ultrasonic transducer acts as both an ultrasonic transmitter and an ultrasonic receiver, so it cannot transmit and receive ultrasonic waves at the same time. A period of time must pass after the ultrasound is emitted before the returning sound waves can be processed. If the obstacle is too close (<>

The third is that the detection beam angle is too large and the directionality is poor, often only the distance information of the target can be obtained, the boundary information of the target cannot be accurately provided, and the stability of a single sensor is not ideal. In practical applications, other sensors are often used to compensate, or multi-sensor fusion technology is used to improve detection accuracy.

Fourth, due to the influence of ultrasonic waves by environmental temperature, humidity and other conditions, and the inherent wide beam angle of ultrasonic waves, the error between the measured value and the actual value of the ultrasonic sensor is relatively large when measuring the distance.

(2) Capacitive coupling sensor

Capacitively coupled sensors detect the presence of obstacles by changing the capacitance when an object approaches the sensor. These sensors are stable, reliable and durable. The disadvantage is that due to the low resolution of the sensor, the dimensionality of the object cannot be resolved within the range of its measurement. The robot must assume that the obstacle is very large when dealing with it. For example, if the distance of the obstacle is 2cm, it is considered to be an object of 20∽30cm to handle, which greatly limits the space for the robot arm to operate.

(3) Eddy current sensor

Eddy current sensors induce eddy currents to surrounding targets by emitting high-frequency changing electromagnetic fields. The magnitude of the eddy current is related to the distance between the sensor and the target object, and the magnetic field generated by the eddy current is opposite to that of the sensor. The superposition of the two magnetic fields reduces the inductance and impedance of the sensor. Using an appropriate circuit to convert the change in impedance into a change in voltage, the distance to the target object can be calculated.

Eddy-current sensors are small in size, high in reliability, and cheap in price. They can not only be used as proximity sensors to detect the presence of obstacles and object distances, but also to detect forces, moments or pressures using appropriate methods. The measurement accuracy is relatively high, and it can detect a small displacement of 0.02mm, and the measurement is also directional. However, the disadvantage of this kind of sensor is that the action distance is short (generally no more than 13mm). In addition, this sensor is only suitable for detection of obstacles that are solid conductors.

(4) Infrared sensor

Infrared sensor is a relatively effective proximity sensor. It is often used by scholars at home and abroad in the obstacle avoidance system of multi-joint robots. It is used to form a large-area robot “sensitive skin”, which covers the surface of the robot arm and can detect the operation process of the robot arm. various objects in. The light emitted by the sensor has a wavelength in the range of a few hundred nanometers, and is a short-wavelength electromagnetic wave. Infrared sensors have the following characteristics: no interference from electromagnetic waves, no noise source, and non-contact measurement can be achieved. In addition, infrared rays (referring to the middle and far infrared rays) are not affected by the surrounding visible light, so it can be measured day and night.

Similar to the sonar sensor, the infrared sensor works in the transmit/receive state. This sensor emits infrared light from the same emission source and uses two photodetectors to measure the amount of light reflected back. Because these instruments measure light differently, they are greatly affected by the environment, the color of the object, the direction, and the surrounding light can all contribute to measurement errors. But since the emitted light is light and not sound, it can be hoped that many infrared sensor measurements can be obtained in a relatively short period of time. The distance measurement range is relatively close, roughly within 30cm.

3. Sensor selection strategy

The selection of sensors is directly related to the amount of information collected by the multi-joint robot from the surrounding environment. Therefore, there are two different methods for selecting the type and quantity of sensors for the robot obstacle avoidance system: the selection method based on the optimization principle of the environment and the method based on the task selection.

(1) Environment-based optimization principle selection method: pre-selection in the design stage and real-time selection suitable for changes in the environment and system state, the former gives the relationship between the appropriate number of sensors and operating speed, which can determine the multi-sensor avoidance The optimal arrangement of the sensor units in the obstacle system, the latter uses any prior object information to determine the location of the sensor through a Bayesian method, so that the sensor has the least uncertainty about the obstacle object assumption.

(2) Task-based selection method: The main idea of ​​this method is based on the task of avoiding obstacles. The process of completing the task is divided into several sections according to time and perception range, that is, the task is decomposed, and the sensor information required for each stage is reasonable. Choose the type and number of sensors appropriately.

3. Information fusion of sensors

In the obstacle avoidance system of intelligent robots, because the function of any sensor is limited, if necessary, multiple sensors should be integrated together to fuse multiple sensor information, so that the characteristics of the external environment can be more accurately and comprehensively reflected. Provide the correct basis for obstacle avoidance. Information fusion technology can increase the complementarity of various sensor information, adaptability to environmental changes, and improve the accuracy of decision-making.

The basic purpose of multi-sensor data fusion is to obtain more information than each single sensor through comprehensive processing of multi-sensor data. It can also be understood as an intelligent synthesis of the original information of multiple sensors to derive new meaningful information. The value of this information is much higher than that obtained by a single sensor, and it is conducive to judgment and decision-making. Therefore, in recent years, the multi-sensor information fusion technology system has been increasingly used in the robot’s obstacle avoidance system, and good results can be achieved through experiments.

1. Sensor data fusion method

In the multi-sensor robot obstacle avoidance system, the environmental information provided by each information source has a certain degree of uncertainty. In addition, due to the large number of sensors and most of them are nonlinear, it is necessary to carry out good global optimization and control, and the processing capacity is large. In the face of many discrete data, high correlation, non-linearization of input information and high reliability of fusion results, traditional data fusion methods (weighted average method, Bayesian estimation method, Dempster-Shafer evidence inference method, etc.) meet the requirements well. For the multi-joint robot obstacle avoidance system, the Kalman filter method, the production rule, and the fuzzy logic artificial neural network method are usually used to obtain a more reliable, unified and accurate description of the environment, which is convenient for judgment and decision-making.

(1) Kalman filter is used for real-time fusion of dynamic low-level redundant multi-sensor data. This method uses the statistical characteristics of the measurement model to recursively determine the optimal fusion data estimation in the statistical sense. Since the robot obstacle avoidance system has a linear dynamic model, and the system noise and sensor noise are white noise models with Gaussian distribution, Kalman filtering provides the only optimal estimate in statistical significance for fusing multi-sensor data.

Applied to the multi-sensor information processing of the robot obstacle avoidance system, domestic and foreign scholars often choose the joint Kalman filter method. The basic idea is to use a set of filter modules running in parallel, each module only processes a specific sensor Information. In addition, a “main filter” is used to fuse the information from all local filters. The obvious advantage of this structure is that the calculation amount is evenly distributed in each parallel filter, and the calculation burden of the main filter is not large; it has a variety of redundant information, and can provide strong fault tolerance through appropriate reconstruction algorithm design.

(2) The production rules can establish a natural scene expert system. According to the detection data of multiple sensors, symbols are used to represent the environmental characteristics, which can more comprehensively reflect the surrounding information of the obstacle avoidance system and prepare for the path planning of the robot.

(3) The fuzzy logic method is to systematically reflect the uncertainty of the multi-sensor data fusion process in the robot obstacle avoidance system with a model that simulates human thinking habits, and complete the data fusion through fuzzy reasoning to obtain the expected effect.

(4) The artificial neural network method is an information processing method that imitates the biological nervous system. It performs network learning through a teacher or no teacher self-learning algorithm. Once the learning is completed, the neural network can be based on the network weight matrix and network topology structure. Based on the feature information stored in the form, a model structure for decision-making thinking is obtained based on this neural network. By synthesizing the information from various sensors in the system, accurate and reliable information that a single sensor cannot provide can be extracted from it. A very efficient method for processing multi-sensor information in an interactive situation.

This method is applied to the multi-sensor information processing of the robot obstacle avoidance system. It mainly obtains the environmental information through the sensor at the operation site. The filtering and preprocessing module corrects and digitizes the sensing information, and then acts as the corresponding neural network fusion processor after being judged by the safety mechanism. The input source of the knowledge database is used as an auxiliary decision-making tool for the selection of the neural network fusion device and the knowledge source. The application program receives the fusion result, adopts the corresponding control strategy, and sends the control command to the robot driving device. In this way, as much environmental information as possible of the actual operation site can be obtained quickly and accurately, so as to effectively complete the multi-sensor information processing.

2. Sensor information processing

Due to the variety and quantity of sensors used in the robot obstacle avoidance system, the information processing is more complicated. The signal processing methods applied in this system mainly include wavelet analysis method, neural network method, genetic algorithm and immune algorithm.

(1) Wavelet analysis method

The basic idea of ​​wavelet transform is to use a family of wavelet base functions to represent or approximate the signal, which solves the contradiction between time and frequency resolution and is suitable for local analysis of time-varying signals.

Wavelet transform is a new signal processing method. In recent years, wavelet analysis has been applied to the detection and analysis of real-time sensor signals collected by robot obstacle avoidance systems. This can effectively improve the reliability of the sampled data in the robot obstacle avoidance system, thereby improving the control accuracy of the obstacle avoidance system. In addition, it has a data compression function, which can save storage space and improve operation speed by compressing a large number of sensor signals in this system.

(2) Neural network method

Neural network is a nonlinear function approximation method that does not need to select a basis function system. The robot obstacle avoidance system uses the highly nonlinear description ability of the neural network, and uses this ability to model the multi-sensor of the system. Using the BP algorithm (error back propagation algorithm), the sensor output signal can be filtered and de-noised. And the signal recognition of the sensor, so that the output signal of the sensor more accurately reflects the external environment information, and prepares for the robot’s path planning algorithm.

The characteristics of this method are: no detailed knowledge of the mechanism is required, and the incompleteness of mathematical modeling is avoided; the processing of sensor signals is realized by software, which is convenient and flexible, and has strong applicability, eliminating the need for hardware circuits.

(3) Genetic algorithm

Genetic algorithm is a global optimization adaptive probabilistic search algorithm proposed in accordance with the principle of “survival of the fittest and survival of the fittest” in nature. The genetic algorithm generates a new generation of population by applying a series of operations such as selection, hybridization, and mutation to the current population, and gradually makes the population evolve to the optimal solution state.

The genetic algorithm is applied to the sensor signal processing of the robot obstacle avoidance system. First, the actual sensor signal is evenly sampled N times in a sampling period and sent to the computer, and several groups of data are randomly selected as the initial group. Then the three operations of selection, hybridization and mutation are carried out cyclically until the given required voltage value is reached. In the robot obstacle avoidance system, the simple amplifying circuit and genetic algorithm software can accurately restore the sensing signal in the case of multiple sensing signals, and improve the measurement accuracy in the sensor information processing.

(4) Immune algorithm

Immune algorithm is a computational method based on simulated organisms. The algorithm simulates the interaction of antibodies and antigens in the immune system. As well as the elimination of antigens by antibodies to achieve digital signal processing.

In recent years, the immune algorithm has also been applied to the sensor signal processing of the robot obstacle avoidance system. This method simulates the action mechanism of the immune system. By processing the complex and large number of sensor signals in this system, the overlapping sensor signals can be obtained. A single set of sensor information, running fast, which can reduce the time for the computer to process the sensor information.

3. Sensor fault diagnosis

The implementation of sensor fault diagnosis can ensure that the diagnosis system can obtain real-time and accurate information, avoid negative effects caused by wrong information, and ensure the correctness of data. Therefore, sensor fault diagnosis is an important guarantee for the system to avoid obstacles in real time. The methods used in the sensor fault diagnosis of the robot obstacle avoidance system mainly include the following aspects:

(1) Fuzzy diagnosis method

Fuzzy diagnosis method is a kind of fault diagnosis method which is based on fuzzy mathematics and carries out state identification, reasoning and decision-making according to the fuzzy state of the sensor of the system.

The advantage of the fuzzy fault diagnosis method is that it can make full use of the expert experience, consider the fuzziness of the fault state and expert experience, and make the diagnosis result more reasonable. It is applied on the computer, and the accuracy rate is also high. It is often used in robot obstacle avoidance systems by scholars at home and abroad to diagnose sensor output results. However, the fuzzy fault diagnosis method also has its imperfect aspects, such as the selection of membership functions and the application of various diagnostic rules.

(2) Discrete wavelet network method

The discrete wavelet network method is to use the wavelet network to diagnose the sensor object in the obstacle avoidance system. When the sensor object has no mutation, the difference between the output of the wavelet network and the output of the sensor object in the diagnosis obstacle avoidance system is small. When the sensor has a mutation, the wavelet The difference between the output of the network and the output of the sensor object in the diagnostic obstacle avoidance system is large, according to which the variance can be used to detect the fault. The method has high flexibility, strong ability to overcome noise, low requirements for input signals, and no mathematical model of the object. Disadvantages: In large scale, due to the large width of the filter time domain, there will be a certain delay in detection.

(3) Artificial neural network diagnosis method

The artificial neural network method has been applied in the field of sensor fault diagnosis in the robot obstacle avoidance system in recent years. Artificial neural network is a network with a parallel processing mechanism, and it can acquire external knowledge through learning. The knowledge distribution stores the connection weights between each neuron. It can complete the complex mapping from input mode to output mode and has fault tolerance. Strong and fast running characteristics.

The method of using the neural network method to diagnose the fault of the robot obstacle avoidance system is as follows: ① Select the key sensor output in the system as the input variable of the neural network, and specify the output variable value of the network; ② Select the appropriate type and structure of the neural network; ③ According to the Select the historical data of the input and output signals, train the network offline, and obtain the weights or thresholds of the network; ④ Apply the previously selected input and output data to the network online, and the network output can give diagnostic results.

The advantage of this method is that it does not require an accurate mathematical model, and the process data can be directly used to solve the fault diagnosis problem of the robot obstacle avoidance system. However, there are still some problems in this method, such as how to select the network structure. In addition, in the process of diagnosis, self-learning and self-diagnosis are often used. Therefore, how to introduce the unsupervised training algorithm into the field of sensor fault diagnosis is also a direction that has been discussed.

4. Conclusion

The real-time obstacle avoidance problem of intelligent multi-joint robots is the key and difficult problem in the field of robot research. In the process of obstacle avoidance, it is often faced with an environment that cannot be known in advance, unpredictable or dynamically changing. The means for robots to perceive the environment are usually incomplete. The data given by sensors is incomplete, discontinuous and unreliable. There are still many problems in the algorithm of sensor information fusion. However, due to the rapid development of sensor technology, the in-depth research of neural networks, fuzzy control theory and other disciplines, and the application of sensor information processing methods, it provides a possibility for the final solution of obstacle avoidance problems, but for complex applications, it is still not enough. Satisfied, so the existing problems are also the research direction in this field.

(1) Sensor fusion technology has been introduced into the research of robot obstacle avoidance in recent years, and has achieved good results. For some high-precision multi-joint robot obstacle avoidance systems, it is difficult to use conventional sensors to meet the performance indicators. Therefore, the development of Novel sensors or constructing sensor arrays according to a certain fusion strategy to make up for the defects of a single sensor will be an important research direction.

(2) Artificial intelligence can make the robot obstacle avoidance system itself have better flexibility and understandability, and at the same time can deal with complex problems. Therefore, various methods of artificial intelligence will be used in the future data fusion technology based on knowledge. The composition of multi-sensor data fusion will still be one of its research trends.

(3) In order to realize the multi-sensor data fusion of the robot obstacle avoidance system, the processor structure will be developed towards the parallel structure, including the parallel structure of the sensor function and the parallel structure of the algorithm function.

(4) In an intelligent system, the use of a single intelligent control method often cannot achieve satisfactory results. Only by comprehensively adopting conventional control methods and intelligent control methods can good results be achieved. Neural networks and fuzzy reasoning are two important tools in obstacle avoidance research, but the research on the integrity of neural network sample sets has not yet made a breakthrough, and it is obviously not advisable to use every point in the event space as a learning sample of the network; fuzzy logic reasoning It focuses on the selection of fuzzy rules, but some rules are difficult to describe formally, or a large number of rules must be used to increase the amount of calculation, which deviates from the original intention of fuzzy logic applications. , using neural network technology to realize the rapid identification and classification of the current perception environment by robots, and then using fuzzy logic technology to achieve a new method of safe obstacle avoidance, it will be a potential research direction.

(5) In the research of centralized multi-sensor system, simulation technology and real-time control technology should be combined, and an integrated development environment should be established to process sensor signals. For distributed sensor systems, a communication-based implementation method should be sought to process sensor signals, which is one of the development directions of sensor systems in the future.

(6) The more advanced the obstacle avoidance system of the robot, the more sensors, the more complex the information processing, and the problem of multi-rate sampling will be encountered. But the existing mature computer control theory involves single-rate sampling, that is, it is assumed that all A/D and D/A channels in the system work at the same sampling rate. To fill this gap, it is necessary to study the modeling, analysis and design methods of multi-rate sampling control systems. Therefore, the research on multi-sensor multi-rate sampling control system for robots is one of the future development directions of sensor systems.

(7) The multi-joint robot obstacle avoidance system is a complex intelligent system. Therefore, in practical applications, various functions must be considered comprehensively. This is an interdisciplinary subject involving mechanics, electronics, computers, automation, physics, etc. The emergence of any new technology may bring breakthroughs in research in this field. Therefore, at the same time as robotics research, we must pay close attention to the development of related disciplines.

Responsible editor: gt

Leave a Reply

Your email address will not be published.