A Priority-Based Multimodal Data Transmission Mechanism for Indoor Autonomous Mobile Robots over Wi-Fi

Indoor autonomous driving systems are used to automate tasks in environments such as warehouses, manufacturing facilities, and service environments by perceiving the surrounding environment and autonomously navigating to a destination. In particular, Autonomous Mobile Robots (AMRs) perform autonomous navigation in indoor environments by using various sensors to perceive their surroundings, estimate their position, and plan driving paths. These autonomous driving systems utilize various sensors, such as LiDAR, Camera, and Radar, to collect different types of information about the surrounding environment and comprehensively perceive the driving environment based on the collected data. LiDAR uses lasers to measure the distance to surrounding objects and environments and can collect three-dimensional spatial information in the form of point clouds. This enables the identification of spatial information, such as the location and shape of surrounding obstacles and terrain. In indoor autonomous mobile robots, LiDAR-based Simultaneous Localization and Mapping (SLAM) techniques are widely used for stable localization and spatial perception. Cameras provide visual information, such as the color, shape, and texture of objects based on image data, and in recent years, numerous studies have actively investigated the use of cameras for deep learning-based object detection and environmental perception. Radar can measure the distance and velocity of objects and is relatively less affected by lighting and certain environmental conditions [1]. Since each sensor provides different characteristics and types of information, AMRs can comprehensively perceive their surrounding environment by utilizing data collected from multiple sensors. In recent years, research on developing and validating autonomous driving and robotic systems using virtual environments, in addition to real-world environments, has also been actively conducted. In particular, robotic simulation environments such as NVIDIA Isaac Sim can be used to construct virtual indoor spaces and generate sensor data required for AMR autonomous navigation by applying various sensors. In addition, various driving scenarios that may occur in real-world environments can be reproduced and repeatedly tested in virtual environments, making them useful research environments for the development and validation of sensor-based robotic systems. As described above, virtual simulation environments can be utilized for research on sensor data generation and autonomous driving system validation for indoor AMRs. They can also be applied to studies on communication and data processing based on data generated from various sensors. Meanwhile, in indoor autonomous driving and robotic systems, sensor data perception and computation can be performed entirely within the vehicle or robot, or alternatively, these tasks can be performed on a server or edge server. In particular, to compensate for the limited computational resources of AMRs, a system architecture can be utilized in which sensor data is transmitted to an external server or edge server for tasks such as image processing, object detection, and environmental perception. In such an architecture, large volumes of sensor data collected by the AMR need to be transmitted wirelessly to the server or edge server. Various wireless communication interfaces, such as Wi-Fi, LoRa, and LTE, can be used for data transmission between an indoor AMR and a server or edge server. Among these, Wi-Fi can provide a relatively high data transmission rate in indoor environments and can establish a wireless network without relying on a separate cellular network, making it suitable for transmitting large volumes of sensor data to a server or edge server. In contrast, LoRa is suitable for low-power and long-range communication but has limitations in transmitting large volumes of sensor data, while cellular communication technologies such as LTE require an architecture that utilizes a cellular network. Therefore, this study focuses on the Wi-Fi communication interface, which can be utilized to establish an indoor wireless network, considering an environment in which large-volume sensor data, such as LiDAR point clouds and Camera images generated by an indoor AMR, are transmitted to a server or edge server. Data generated by different sensors varies in size and characteristics depending on the type of data, and transmitting data from multiple sensors simultaneously over the same Wi-Fi network can increase network traffic. In particular, when large volumes of LiDAR point cloud and Camera image data are transmitted simultaneously, packet loss and transmission delays may occur due to limited wireless bandwidth [2]. As communication delays increase, sensor data may not be delivered to the server or edge server in a timely manner, which can affect the environmental perception and navigation processes of the AMR [3]. Therefore, a data transmission method is required to efficiently deliver large volumes of data generated by multiple sensors in a bandwidth-limited Wi-Fi communication environment. To address this issue, studies have been conducted on determining transmission priorities by considering the importance of data in vehicle and robotic communication environments. Giordani et al. proposed a framework that evaluates the Value of Information (VoI) and determines transmission priorities according to the importance of information [4]. Fang et al. proposed a Priority-Aware Cooperative Perception (PACP) framework that determines priorities based on the correlation of perception information between vehicles [5]. Sato et al. proposed a priority-based transmission method that considers the spatial importance of LiDAR point clouds [6]. Chen et al. demonstrated that prioritizing LiDAR data can contribute to improving the safety of autonomous driving systems through a pipeline that performs LiDAR-based object clustering before Camera inference [7]. Meanwhile, Butt et al. analyzed the importance of multimodal sensor fusion involving Radar, LiDAR, and Camera, as well as vehicle communication infrastructure, in autonomous driving systems [8]. Kocić et al. showed that sensor fusion of Camera, Radar, and LiDAR can contribute to improving environmental perception performance. In recent years, object and environmental perception methods using Camera-based deep learning techniques have been actively studied. However, in indoor AMR environments, LiDAR-based SLAM is frequently utilized to achieve stable spatial perception and localization. Unlike image-based information, LiDAR directly measures the distance between the sensor and surrounding objects and can represent the structure of the surrounding space in three dimensions. This enables the AMR to identify the terrain and locations of obstacles within its operating space and obtain geometric information such as the height and shape of structures and the spatial arrangement of the surrounding environment. Since this information can be used to estimate the current position of the AMR and determine navigable areas and driving paths, it can be regarded as essential information for identifying the basic structure of the surrounding space during the initial stage of navigation. Therefore, in a bandwidth-limited communication environment, rather than obtaining all sensor data simultaneously, it is necessary to acquire LiDAR data first, as it provides the information required to determine the AMR's movement path and the basic structure of the surrounding space. Accordingly, this study assigns LiDAR data to Priority 1 and transmits it first, followed by Camera and Radar data after the reception of LiDAR data has been completed. Here, the priority assigned to LiDAR does not imply that it is absolutely more important than Camera or Radar. Rather, it refers to the transmission order from a communication perspective, intended to first secure the spatial and localization information required to identify the initial driving environment of the indoor AMR. However, existing studies have primarily focused on determining transmission priorities based on the importance and information value of messages or individual data items. Relatively few studies have directly controlled the transmission order among sensor modalities by considering the functional characteristics of each sensor during the environmental perception process of indoor AMRs. In addition, in environments where multimodal sensor data is transmitted simultaneously, it is necessary to determine which data should be acquired first by considering not only the importance of the sensor data but also the limited bandwidth of Wi-Fi. Therefore, a data transmission method is required that can prioritize the acquisition of basic spatial and environmental information necessary for AMR navigation, even when communication conditions deteriorate. This paper proposes a priority-based data transmission mechanism for efficiently transmitting multimodal sensor data from indoor Autonomous Mobile Robots (AMRs) in a Wi-Fi-based communication environment. The proposed mechanism establishes transmission priorities for sensor data by considering the environmental perception process of indoor AMRs and the functional roles of individual sensors. First, LiDAR data, which provides distance and spatial information about the surrounding environment, is assigned Priority 1 and transmitted first. After the transmission and reception of LiDAR data are completed, Camera and Radar data are assigned Priority 2 and transmitted in parallel. This approach manages the network load generated during the simultaneous transmission of multiple sensor data streams and enables the basic spatial and environmental information of the AMR to be acquired before the reception of all sensor data is completed. The main contributions of this paper are as follows. First, we propose a data transmission model that defines the transmission priorities of LiDAR, Camera, and Radar data by considering the environmental perception and localization processes of indoor AMRs and the functional characteristics of each sensor. Second, we design a priority-based multimodal data transmission mechanism that transmits LiDAR data first, followed by the parallel transmission of Camera and Radar data, thereby managing the transmission load of multimodal sensor data in a Wi-Fi environment. Third, we present a structure that prioritizes the acquisition of LiDAR data even in a bandwidth-limited Wi-Fi environment, enabling the basic spatial and environmental information of the AMR to be acquired before the transmission of all sensor data is completed. The remainder of this paper is organized as follows. Section II describes the system model for transmitting multimodal sensor data from indoor AMRs and the method for assigning priorities to sensor data. Section III describes the proposed Wi-Fi-based priority data transmission mechanism and the data transmission procedure. Section IV presents the indoor AMR simulation environment configured using NVIDIA Isaac Sim and the experimental results, and analyzes the performance of the proposed method. Finally, Section V presents the conclusions of this study and discusses directions for future research.

II. SYSTEM MODEL This study considers a communication environment in which multimodal sensor data generated by an indoor AMR is transmitted to a server or edge server via Wi-Fi. The proposed system consists of a sensor data collection unit on the AMR, a Wi-Fi communication network, and a receiving system. The AMR uses LiDAR, Camera, and Radar to collect different types of information about the surrounding environment, and the collected data is transmitted to the receiving system through the Wi-Fi network. In an indoor AMR environment, sensor data perception and computation can be performed within the AMR. However, considering the limitations of computational resources, an architecture in which some perception and computation tasks are performed on a server or edge server can also be considered. In such an architecture, sensor data generated by the AMR must be transmitted to the server or edge server, making the bandwidth and data transmission delay of the wireless network important factors. In a conventional multimodal sensor data transmission method, data generated by multiple sensors may be transmitted simultaneously using the same Wi-Fi communication resources. However, when multiple sensor data streams, including large volumes of LiDAR and Camera data, are transmitted simultaneously in a bandwidth-limited Wi-Fi environment, instantaneous network traffic may increase. When the traffic exceeds the processing capacity of the network, packet loss and transmission delays may occur, which can delay the acquisition of necessary environmental information at the server or edge server. Therefore, this study controls the transmission order of sensor data by considering the functional characteristics of each sensor and the environmental perception process of the indoor AMR. After collecting LiDAR, Camera, and Radar data at the AMR, the transmission priority of each data type is determined, and LiDAR data, which has the highest priority, is transmitted first. Once the receiving system completes the reception of the LiDAR data, it can use the data to construct basic distance and spatial information about the environment surrounding the AMR. Subsequently, Camera and Radar data are additionally received to supplement visual information and object distance and velocity information that are difficult to obtain from LiDAR data alone. The priority control and transmission/reception flow of multimodal sensor data for the Wi-Fi-based indoor AMR is illustrated in Figure 2. B. Multimodal Data Priority An indoor AMR must continuously perceive environmental information, such as surrounding obstacles, structures, and moving objects during navigation. LiDAR, Camera, and Radar have different sensing characteristics and provide different types of information, including distance, spatial structure, color, texture, and velocity. Since these sensor data provide complementary information, they can be jointly utilized in the environmental perception process of an AMR. LiDAR can directly measure the distance to surrounding objects, thereby enabling the spatial structure of the surrounding environment to be identified. In particular, indoor AMRs can utilize LiDAR-based SLAM to generate maps of the surrounding environment and estimate the position of the AMR. Therefore, LiDAR data can be regarded as data that can be used to acquire basic spatial information about the surrounding environment and estimate the position of the AMR. Camera provides visual information, such as the color, shape, and texture of objects, based on image data. In recent years, it has also been widely used to extract various types of environmental information through deep learning-based object detection and classification techniques. Radar can measure the distance and velocity of objects and provide complementary information to that provided by Camera and LiDAR. In this study, the order in which sensor data is acquired in a bandwidth-limited Wi-Fi communication environment is defined by considering the functional characteristics of each sensor. The priority defined in this paper does not mean that a particular sensor is absolutely more important than the other sensors. Rather, it represents the transmission order for acquiring the initial spatial and environmental information of an indoor AMR first under limited communication resources. To this end, LiDAR data is assigned Priority 1, while Camera and Radar data are assigned Priority 2. Since LiDAR data provides distance and spatial information about the surrounding environment and can be utilized for spatial perception and localization of indoor AMRs, it is transmitted first to construct the initial environmental information. Subsequently, Camera and Radar data are transmitted to supplement the basic spatial information obtained through LiDAR with visual information as well as distance and velocity information. Accordingly, the basic data transmission structure of the proposed system is configured to transmit LiDAR data first, followed by the parallel transmission of Camera and Radar data. Camera and Radar, which have the same Priority 2, can be transmitted in parallel over the same Wi-Fi communication environment after the transmission and reception of LiDAR data have been completed. With this priority configuration, the system does not need to wait until the transmission of all multimodal data is completed to acquire environmental information. Instead, once the reception of LiDAR data is completed, basic spatial information can be acquired first, after which the environmental information can be supplemented using Camera and Radar data.

A. Priority-based Multimodal Data Transmission Unlike simultaneous sensor data transmission, the proposed mechanism controls the transmission order of multimodal data according to predefined priorities. In a conventional simultaneous transmission method, LiDAR, Camera, and Radar data generated by the AMR are transmitted simultaneously over the same Wi-Fi network, causing each sensor data stream to compete for the same communication resources. As a result, instantaneous network traffic may increase, and packet loss and transmission delays may occur due to limited bandwidth. In the proposed method, after collecting multimodal data at the AMR, the type and assigned Priority of each data stream are identified. First, LiDAR data, which has the highest Priority, is transmitted through Wi-Fi. Once the reception of the LiDAR data is completed, the receiving system uses the data to construct basic distance and spatial information about the environment surrounding the AMR. After the reception of the LiDAR data is completed, Camera and Radar data, which are assigned Priority 2, are transmitted in parallel. Camera data provides visual information about objects, while Radar data provides distance and velocity information about objects, thereby adding complementary environmental information to the basic spatial information obtained from LiDAR data. Finally, the received LiDAR, Camera, and Radar data are combined to construct comprehensive environmental information that can be utilized for environmental perception of the AMR. In this structure, rather than delaying the construction of environmental information until the transmission of all sensor data is completed, LiDAR data is acquired first to construct the initial spatial information, after which Camera and Radar data are added to supplement the environmental information. Therefore, the proposed mechanism reduces the simultaneous transmission of multimodal data in a Wi-Fi environment and enables limited communication resources to be utilized according to predefined priorities. B. Data Transmission Procedure In the proposed mechanism, the transmission target is determined according to the predefined Priority when multimodal data is generated. When LiDAR, Camera, and Radar data are generated by the AMR, the type and transmission status of each data stream are checked. If the LiDAR data is in a transmission-waiting state, it is transmitted first through the Wi-Fi network. The receiving system checks whether the reception of the LiDAR data has been completed. Once the data is successfully received, it is used to construct basic distance and spatial information about the environment surrounding the AMR. Subsequently, Camera and Radar data corresponding to Priority 2 are received in parallel to add visual information and object distance and velocity information to the environmental information obtained from the LiDAR data. Once the reception of all multimodal data is completed, the received LiDAR, Camera, and Radar data are integrated to construct the final environmental information that can be utilized for environmental perception of the AMR. This procedure adjusts the reception order of sensor data according to the environmental perception process, enabling initial environmental perception to be performed before all data has arrived. Even when all sensor data cannot be received immediately due to poor network conditions, prioritizing the acquisition of LiDAR data allows basic information about the surrounding environment and spatial structure to be constructed before the reception of all sensor data is completed. This enables the initial spatial information required for AMR navigation to be acquired preferentially even in a bandwidth-limited Wi-Fi communication environment. The key concept of the proposed mechanism is not to transmit all sensor data at the same time, but to transmit the spatial information required during the initial stage of the environmental perception process of an indoor AMR first. To this end, LiDAR data is transmitted first to acquire basic environmental and spatial information, followed by the parallel transmission of Camera and Radar data to supplement the environmental information. Consequently, in a Wi-Fi-based multimodal data transmission environment, the transmission order of sensor data is configured according to the environmental perception process, enabling the efficient utilization of limited communication resources. To determine the transmission order of sensor data, a priority model is defined by assigning weights to each sensor. In this paper, the weights are configured based on the environmental perception process of autonomous navigation as follows. At (1), wL, wC and wR represent the priority weights assigned to LiDAR, Camera, and Radar data, respectively. The transmission order is determined by transmitting sensor data with higher weights first. Camera and Radar data, which have the same weight, can be transmitted in parallel after the transmission of LiDAR data has been completed. Considering the Wi-Fi bandwidth and the size of the sensor data, the total transmission completion time for the conventional simultaneous transmission method is defined as follows. Let the sizes of the LiDAR, Camera, and Radar data be defined as SL, SC and SR, respectively, and let the available bandwidth of the Wi-Fi network be defined as B. Here, SL, SC and SR denote the sizes of the LiDAR, Camera, and Radar data, respectively, while B denotes the available bandwidth of the Wi-Fi network. At (2), Tsim represents the total transmission completion time of the conventional simultaneous transmission method. In the proposed method, LiDAR data is transmitted first, followed by the parallel transmission of Camera and Radar data. Therefore, the total transmission completion time is given as follows. At (3) , Tprop represents the total transmission completion time of the proposed method, T represents the transmission time of the sensor data, S represents the size of the sensor data, and B represents the available bandwidth of the Wi-Fi network. Specifically, SL, SC and SR represent the sizes of the LiDAR, Camera, and Radar data, respectively.

In this paper, we proposed a priority-based multimodal data transmission mechanism for efficiently transmitting multimodal data from an indoor Autonomous Mobile Robot (AMR) in a Wi-Fi-based communication environment. The proposed mechanism assigns LiDAR data to Priority 1 by considering the spatial information required for environmental perception and localization of the indoor AMR. After the transmission and reception of LiDAR data are completed, Camera and Radar data are assigned to Priority 2 and transmitted in parallel. This enables the AMR to preferentially acquire LiDAR data in an environment where multiple sensor data streams are transmitted simultaneously, allowing basic spatial and environmental information to be obtained before the transmission of all sensor data is completed. In addition, this study presents a structure for controlling the transmission order of sensor data under limited wireless bandwidth, considering an indoor Wi-Fi communication environment in which sensor data must be transmitted between an AMR and a server or edge server. In particular, the proposed mechanism supports the acquisition of initial spatial information by transmitting LiDAR data first, even when network conditions deteriorate or communication resources are limited. As future work, the proposed method will be compared with the conventional simultaneous transmission method for multimodal data in real or simulation-based Wi-Fi communication environments. Quantitative measurements will be conducted using metrics such as Packet Loss Rate, Transmission Latency, LiDAR data acquisition time, and total data transmission completion time. Based on these measurements, the performance of the proposed mechanism will be analyzed under varying network conditions and sensor data sizes. In addition, priority assignment and data transmission methods will be further investigated to support the application of the proposed mechanism to various indoor AMR environments.

This research was supported by the 2026 Kyungpook National University Research Support Program. This research was supported by the Open Source Commercialization Support Program for New Industries (PJT-26-040381), funded by the Ministry of Science and ICT of the Republic of Korea.