Abstract
In the era of artificial intelligence, children's service robots will surely become one of the conventional hardware. This paper combines artificial intelligence technology and the space algorithm to design the child service robot system, improves the situational space perception algorithm, and applies the algorithm to the robot system. Moreover, this paper constructs the system structure of the children's service robot according to the actual needs and obtains the system function modules. In addition, this paper constructs a child service robot based on context perception and verifies the performance of the robot system. Finally, this paper combines the simulation experiment to carry out the interactive effect of the robot system and the service effect to children and investigates the satisfaction of parents. Through experimental research, it can be seen that the child service robot system based on situational awareness constructed in this paper has good service effects, can effectively take care of children, and has high parent satisfaction.
1. Introduction
In recent years, with the development of artificial intelligence technologies such as big data, deep learning, cloud computing, sensors, and machine vision, more and more intelligent products endowed with AI technology have penetrated into various fields of life, and the intelligence of traditional products has become an inevitable trend [1]. The field of children’s products is no exception. Traditional static children’s products can no longer meet the complex, dynamic, and emotional scenes in life. Therefore, various technology companies and traditional manufacturers actively introduce new technologies and develop intelligent and emotional products that truly meet the needs of children [2]. As one of the important branches of the robotics field, the companion robot for children is favored by the parents of the post-80s and post-90s due to its own emotional, interactive, and intelligent advantages. The high pressure of life, the fast pace, and the less time to spend with children are the main pain points of parents nowadays. However, the emergence of child companion robots has alleviated this problem to some extent. Children's companion robots can provide children with enlightenment education and scene-based entertainment companions through technologies such as big data computing, deep learning, and sensor capture. Moreover, it can effectively strengthen the interaction between children and the outside world, which is conducive to early childhood physical and mental development [3].
In recent years, with economic development and changes in lifestyles, people have put forward higher requirements for the quality of life. On the one hand, as the pace of life accelerates and work pressure increases, people hope to be relieved from the complicated housework; on the other hand, the problem of aging population has become prominent, and the pressure on social pensions is increasing. In addition, the number of people with disabilities is gradually increasing, and the elderly and the disabled need more care. The development of service robots has brought new technical ideas to solve the above problems, and can play a huge role in improving people's quality of life. Therefore, there has been an upsurge in the development of service robots at home and abroad. As an inevitable extension of the modern information technology revolution, service robots will follow the path of smartphones and PCs, with diversified application scenarios, fast updating, and very broad market prospects. After the realization of personal robots (PR) in the future, the global market capacity will exceed 3 billion units, and a service robot market of 100 billion US dollars can be born, that may reach to an unimaginable scale.
As a typical representative of service robots, escort robots are in a rapid development stage. People have higher and higher functional requirements for escort robots. However, traditional escort robots are designed to operate in a stand-alone mode, which has complex software and hardware systems, expensive construction, and difficult development and maintenance. Skill sharing has seriously hindered its promotion and application. In recent years, technologies such as cloud computing, big data, and virtualization have continuously made breakthroughs, which have promoted the informatization and modularization of the science and technology field, thus forming a new concept “cloud service”, which has high resource utilization and data processing capabilities It has the advantages of strong, high fault tolerance, and low cost. It can coordinate management and scheduling of cloud resources, provide powerful computing capabilities, and provide users with on-demand services. It is widely used in aerospace, medical, financial, and education fields. Based on the intensive computing power of cloud computing, big data, and the analysis and processing capabilities of massive data, the escort robot body function is virtualized to the cloud, and the corresponding cloud service is constructed. While simplifying the software and hardware structure of the robot body, it also reduces the cost of the robot itself., and expand its function beyond the limitations of its physical body, thus realizing the real “good quality and low price.” Therefore, the combination of escort robots and cloud technology is an inevitable trend in the development of service robots in the future.
This article combines intelligent technology to construct a children's service robot based on situational perception, combines intelligent space technology to improve the immersion that children’s service robot brings to children, and combines experimental research to verify the performance of the system.
2. Related Work
The literature [4] has conducted sufficient research on home service robots in the intelligent era. Firstly, it conducted in-depth research from five aspects: the concept of service robot, the control mode of the product, the movement mode of the product, the functional attribute of the product, and the structural design. Then, it divided user research into three parts: children's physical characteristics, behavioral characteristics, and psychological characteristics. Finally, from the user’s point of view, it took the principle of humanity and modularity as the guiding principles for the design of the child care robot. The literature [5] took children’s robots as the research object, used perceptual engineering methods to study its modeling design elements, and established a quantitative model between perceptual images and children's robot modeling design elements. The literature [6] analyzed and studied children’s behavioral characteristics and emotional characteristics, and proposed that the study of children’s groups should pay special attention to the catharsis of negative emotions in children's emotions. Moreover, it illustrated the importance of emotional catharsis to the growth of children and focused the research on the problems of loneliness and negative emotions encountered in the process of children's growth. In addition, it took emotional design as a guiding principle and combined the existing theoretical basis of children’s robot design to innovate its shape and function, which reduces the distance between the robot and the user. The literature [7] put forward the concept and the design method of remote parent-child interaction from the perspective of parent-child games based on user research, so that parents can remotely accompany and communicate with children and communicate emotionally. Child robots are an auxiliary tool for parents to accompany their children in education.
Literature [8] took the emotional design as the theoretical basis, analyzes, and compares the existing products on the market in terms of the design elements, and uses the three levels in the Norman emotional design to target the children’s emotional needs. The innovative design of the product is divided into four levels: form, color, function, and action, and carries out innovative design of the product shape.
Literature [9] proposed the characteristics of children’s products in the 020 mode and introduced the commercialized 020 mode into the design of children’s interactive products. Literature [10] proposed to explain the app interaction strategy of early childhood education in the three parts of information architecture, information presentation, and information efficiency. Literature [11] proposed that children’s interaction problems should be based on children’s spiritual and behavioral needs to meet children’s aesthetic requirements for interfaces. Literature [12] pointed out the importance of other sensory experiences in product interactions in addition to visual experience in the research on the interactive interface of children's products and proposed related research results on graphical and sound interactive interfaces.
Literature [13] proposed a brand-new interactive method that uses multiple thermal sensations in parent-child interaction design to achieve wearable remote interaction. Literature [14] established a user-centered interactive evaluation method for children's robots. Literature [15] proposed that using music interaction to communicate with children for emotional communication and parent-child interaction is superior to other methods.
Literature [16] proposed that when children are playing games with application software, they should actively help children understand the principles of freedom and flexibility of the extensibility of the objects and interaction methods in the interface. The development of service robots is inseparable from the development of interactive technology. Takeo Igarashi provided children with a simple pen and gesture interaction that can generate a rough 3D model. It is a 3D drawing model based on outline Teddy [17]. MIT multimedia laboratory has developed an interface system Doll Talk [18] that exercises children’s conversational ability. This system can simulate speech recognition by capturing children’s gestures and language information, and then inform children of the conversation content by changing the pitch, thereby guiding children to improve their own narrative. IaTAR is an immersive simulation creation system developed by Pohang University in South Korea. The system emphasizes “what you see is what you get” and can cultivate children's imagination and creativity in space [19]. These interactive technologies play a vital role in the interaction between robots and users.
3. Spatial Algorithm Based on Context Perception
As the most typical adaptive beamforming algorithm, the MV algorithm has a certain degree of contribution to improve the imaging resolution, but the algorithm's robustness are not as good as the traditional nonadaptive DAS algorithm: and if the expected direction vector of the MV algorithm is not accurate, or the estimation of the sample covariance matrix is not accurate, which will greatly affect the imaging effect. In engineering practice, due to various noises and interferences in the detection environment, there are errors between the echo signal of the ultrasonic array and the expected direction vector, which makes it difficult to meet the normal working conditions of the MV adaptive algorithm, which leads to a decrease in the output signal-to-noise ratio. This section mainly analyzes the robust form of the MV algorithm under error conditions, that is, an improved algorithm that overcomes the influence of error factors such as noise and interference to improve the output performance and robustness of the MV algorithm. At present, the common beamforming robust algorithms for the MV algorithm includes diagonal loading method to obtain a more accurate sample covariance matrix, a spatial smoothing method to reduce the strong correlation of ultrasonic signals, and a feature space method to filter out the influence of noise interference. This section will conduct detailed analysis and research on the principles and performance of the above robust algorithms [20].
According to the basic principle of the MV algorithm, the acquisition of the optimal dynamic weight vector involves the determination of the desired direction vector and the inverse operation of the sample covariance matrix . In the actual detection process, the number of ultrasonic echo data samples are limited by the influence of noise interference, so that the estimation error of the sample covariance matrix increases, and it is even irreversible. When there is an error in the expected direction vector, the expected signal will be cancelled, which will cause the image quality to deteriorate. In order to effectively improve the estimation accuracy of the sample covariance matrix, some scholars have proposed the diagonal loading method to correct the error of the estimated sample covariance matrix. A constant is added to the diagonal element of the sample covariance matrix. On the one hand, the loaded covariance matrix can be full-rank invertible, so the problem of too few samples is compensated. On the other hand, it can compensate the weakened desired signal to a certain extent, and produce a robust effect on the error of the desired direction vector. The specific principle of the algorithm is described as follows [21]:
Among them, B represents a unit variance random matrix with a mean value of 0, and represents an error factor, which generally takes an integer value. Then, the estimated sample covariance matrix after diagonal loading compensation can be expressed as
Among them, represents the diagonal loading factor, and I represents the identity matrix. According to the principle of the MV algorithm, the minimum variance criterion after diagonal loading compensation will be transformed into
It can be obtained that the mathematical expression of the optimal weight vector of the MV algorithm after diagonal loading can be modified as
Secondly, although the diagonal loading method can enhance the robustness of the MV algorithm against the number of echo signal samples and the desired direction vector, the selection of the diagonal loading factor A is very critical. If the value of A is too small, the estimated sample covariance matrix is irreversible, and the loading effect will be lost; if the value is too large, the algorithm's anti-interference ability will decrease. To solve this problem, a method to determine the value based on the ultrasonic echo signal is proposed. The specific selection criteria are as follows:
We assume that there is , and the approximate inverse operation of the estimated sample covariance matrix can be performed by the following formula to obtain the optimal weighting vector of the MV algorithm:
Among them, A represents the array flow matrix, and is the diagonal matrix composed of the noise power of the diagonal elements. At the same time, represents the loading factor selection parameter, and N represents the number of array elements. According to formula (5), in order to obtain a more accurate estimated covariance inverse matrix, while making the left half term closer to the true sample covariance matrix, the value of the right half term that mainly causes estimation errors should be reduced to zero. That is, the formulas (6) and (7) are satisfied, respectively [22]:
Among them, there is . According to formula (6) and formula (7), the value range of diagonal loading factor is
Among them, the error factor in formula (8) and the diagonal element of the real sample covariance matrix are unknown items in engineering practice, and the value range of can only be obtained based on the estimated value of the sample covariance matrix.
According to the mean variance characteristics of matrix B, the diagonal elements and error factors of the real sample covariance matrix can be estimated by estimating the mean and standard deviation of the diagonal elements of the sample covariance matrix , as shown in formulas (9) and (10):
Among them, trace(·) represents the trace calculation operation, std(·) represents the standard deviation operation, and diag(·) represents the diagonal element solving operation. According to formulas (9) and (10), the value range of can be obtained as
In summary, by adopting the diagonal loading method with appropriate parameters, the full-rank reversible estimated sample covariance matrix can be obtained, thereby increasing the robustness of the MV algorithm.
In the ultrasonic detection process, the echo signal received by each element of the ultrasonic transducer comes from the reflection or scattering of the object to be detected. Therefore, the received ultrasonic echo signal exhibits a strong correlation, resulting in the cancellation of the expected signal, which seriously affects the quality of ultrasonic imaging. In this regard, the spatial smoothing method is proposed to weaken the strong correlation between the echo signals received by the ultrasonic array element. The basic principle of the method is as follows: first, the echo signals received by the ultrasonic transducer arrays with the same array flow pattern are divided, and several overlapping sub-arrays are obtained. Secondly, the sub-array covariance matrix corresponding to each sub-array is solved in turn and its average value is obtained. Finally, this value is used to replace the original sample covariance matrix for adaptive weighting. The ultrasonic linear transducer array of N elements is taken as an example, and the detailed schematic diagram of the forward and backward spatial smoothing is shown in Figure 1 [23].
It can be seen from Figure 1 that the forward and backward spatial smoothing method uses the translation invariance of the linear array to achieve the uniform division of the sub-arrays. Among them, the forward smoothing method divides all array elements uniformly into N−L+1 overlapping sub-arrays including L array elements in the order from left to right. The first sub-array is taken as an example, the received echo signal (k) after forward smoothing can be expressed as
It can be seen that the subsample covariance matrix corresponding to the lth submatrix after forward smoothing can be expressed as
Among them, represents the expected operation and (·)H represents the conjugate transpose operation. The N−L + 1 subsample covariance matrix obtained by equation (13) is averaged, and the sample covariance matrix smoothed by the forward space can be expressed as
Therefore, the spatial smoothing method weakens the strong correlation of the ultrasonic echo signal by averaging the covariance matrix of all sub-arrays, and the dimension of the covariance matrix is reduced from the number of elements N to the number of elements L of the sub-array. As a result, the complexity of the inversion of the covariance matrix is reduced, and the efficiency of the algorithm is improved. However, the application of the forward spatial smoothing method will cause the loss of the array aperture, which will adversely affect the imaging resolution. Therefore, a forward and backward spatial smoothing method is proposed, and N−L + 1 backward sub-arrays are added to perform backward smoothing on the basis of the former. The Ith sub-array is taken as an example, and the received echo signal after backward smoothing can be represented as
In the same way, the corresponding subsample covariance matrix of the lth submatrix after the backward smoothing process can be represented as
Similar to the principle of formula (14), the sample covariance matrix smoothed by the backward space can be represented as
By taking the arithmetic average of the matrices Rf and Rb, the forward and backward spatial smoothing matrix Rfb can be obtained as
However, although the above method can effectively reduce the loss of the array aperture, it doubles the calculation amount compared to the one-way spatial smoothing method. In order to reduce the computational complexity of the algorithm, an appropriate matrix J can be constructed to directly solve Rfb through Rf. The specific conversion steps are as follows:
Among them, J represents a conversion matrix whose anti-angle element is 1 and the remaining elements are 0, and represents a matrix conjoint operation.
The echo signal received by the ultrasonic transducer contains both the desired signal and the noise signal. If the noise signal cannot be effectively filtered out, it will result in a reduction in the signal-to-noise ratio of the adaptive beamforming output, and seriously affect the imaging quality. In this regard, the feature space method is proposed to suppress the noise content in the echo signal, thereby improving the robustness of the MV adaptive beamforming algorithm. The basic principle of the algorithm is: first, the sample covariance matrix of the echo signal received by the array element is decomposed by the eigenvalue, and it is sorted in descending order according to the size of the obtained eigenvalue. Secondly, the aforementioned covariance matrix is decomposed in the signal subspace and the noise subspace that are orthogonal to each other. Since the noise subspace contains noise and interference signals in the ultrasonic echo signal, a specific threshold can be set according to the feature value, and the signal subspace can be set to the feature space corresponding to the feature value greater than the set threshold. Finally, the weighted vector obtained in the adaptive algorithm is projected to the signal subspace to obtain the desired echo signal, and the signal subspace and the noise subspace are orthogonal to each other to filter out the noise component in the signal. Finally, the optimal adaptive weight vector is obtained. The specific principle of the algorithm is described as follows:
First, the eigenvalue decomposition of the estimated sample covariance matrix is represented by its corresponding signal subspace and noise subspace:
Among them, is the signal subspace, num represents its spatial dimension, and is the noise subspace. At the same time, is the N eigenvectors corresponding to the N eigenvalues of the covariance matrix, and its eigenvalue satisfies . and are diagonal matrices formed by the corresponding eigenvalue elements of the signal subspace and the noise subspace.
Secondly, after the signal subspace corresponding to the covariance matrix is obtained, the optimal weight vector obtained in the adaptive algorithm is projected to the signal subspace, which can filter out the component of the weight vector in the noise subspace. Furthermore, the optimal weighting vector based on the feature space is obtained as
In summary, the application of the feature space method can effectively improve the resolution and contrast of ultrasound imaging, and improve the performance of the adaptive algorithm and the imaging effect. However, the operation of this algorithm involves matrix eigenvalue decomposition, eigenvalue sorting, and matrix inversion operations. The amount of calculation and complexity are high, which seriously affects the operating efficiency of the algorithm.
4. Children’s Service Robot System Based on Context Perception
The essence of interaction design is to improve the user's experience through the communication between individuals and make users feel emotional, so that users can experience the product more conveniently and comfortably, and complete the target tasks. Therefore, interaction design is essentially a subject that focuses on interactive experience. In the process of using products and various services, it is actually interacting with them. The feeling during the use is a kind of interactive experience. The interaction design system contains five elements: human, actions, media, goals, and scenarios, as shown in Figure 2.
Applying the psychological CAPS theory to interaction design is to better interpret the user, and from the user's point of view, it is to respect the user's behavior and habits. In particular, for children, a special user group, their cognitive, psychological, and emotional characteristics are different from those of adults. Therefore, it is more necessary to study the cognitive and emotional characteristics of children from the perspective of this special group of children, and then apply them to the design, so that users can feel the “care” that the product brings to people. Therefore, in the interaction design process, we should pay attention to the discovery of user needs, use the theoretical methods of psychology to collect user information, and apply the CAPS theory to the specific design process of children's interaction design, as shown in Figure 3.
The child’s cognitive and emotional system is further generalized and divided into five subsystems: control system, perception system, expression system, memory system, and thinking system, which correspond to and influence children’s cognitive and emotional units, respectively. Thus, the corresponding relationship between the children's cognitive emotion system and the three parts of the product's interactive design solution is summarized, as shown in Figure 4.
According to the analysis of user needs and the current situation of the product, this paper will study the interactive design of the child escort robot at the content level from three aspects, that is, interaction design based on “accompanying” requirements, interaction design based on “educational” functions, and interaction design based on “interest” requirements. This article summarizes the commonality and individuality of existing products, and the results are shown in Figure 5.
As shown in Figure 6, the hardware facilities of the child service robot should include distance sensors, cameras, screens, LED lights, microphones, speakers, power buttons, and roller skating.
The functions of remote real-time monitoring, remote control, and offline monitoring are to meet the needs of parents. Remote monitoring and remote control mean that parents can operate the robot through media devices to observe the child's dynamics when they are working or on a business trip. The purpose of offline monitoring is to understand what has happened. The method adopted is to upload the dynamic scene in front of the camera to the Baidu cloud disk in the form of pictures. In their leisure time, parents learn about the child's dynamics by browsing pictures, on the other hand, they record the child's growth process and monitor the safety of the home at the same time. Therefore, to meet the needs of parents and children at the same time, the child service robot should have the above six functions: education companion, voice-controlled movement, graphical programming, remote monitoring, remote control, and offline monitoring. In summary, the overall structure design of the child service robot is obtained, as shown in Figure 7.
After constructing a child service robot based on context perception, the performance of the robot system is verified. This paper combines the simulation experiment to carry out the interactive effect of the robot system and the service effect to children, and investigates the satisfaction of parents. The results are shown in Table 1 and Figure 8.
From the above research, it can be seen that the child service robot system based on context perception constructed in this paper has good service effects, can effectively take care of children, and has high parent satisfaction.
5. Conclusion
Solving the problem of left-behind children through service robots is a new solution to social problems. Based on the research results of domestic and foreign scholars, this paper summarizes basic theories such as children's physiological and psychological characteristics and interaction design theory. Then, this paper further studies the needs of the left-behind children on the companion robot through interactive behavior experiments, and further explores the design concepts and methods of the companion robot for children. In addition, based on the above research results, a targeted design plan is made as a practice and test of the design theory. Finally, starting from the existing emotional design theory, through actual case analysis, literature search, and practical research, the children’s emotional needs in life and learning are clarified, and a child companion robot design plan is formed, and the product design and practice of children's companion robots are carried out based on the research conclusions.
Data Availability
Data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The author declares that there are no conflicts of interest regarding this work.