Edge-machine collaboration has the potential to facilitate compute-intensive device pose monitoring for useful resource-constrained mobile augmented reality (MAR) gadgets. In this paper, we devise a 3D map management scheme for edge-assisted MAR, whereby an edge server constructs and updates a 3D map of the physical surroundings through the use of the digicam frames uploaded from an MAR machine, to help native system pose tracking. Our goal is to attenuate the uncertainty of gadget pose tracking by periodically selecting a proper set of uploaded digital camera frames and updating the 3D map. To cope with the dynamics of the uplink knowledge price and the user’s pose, iTagPro key finder we formulate a Bayes-adaptive Markov choice course of downside and suggest a digital twin (DT)-based method to solve the issue. First, a DT is designed as a knowledge model to seize the time-varying uplink information price, iTagPro official thereby supporting 3D map management. Second, utilizing in depth generated information provided by the DT, a model-based reinforcement studying algorithm is developed to manage the 3D map whereas adapting to these dynamics.

Numerical results show that the designed DT outperforms Markov models in precisely capturing the time-varying uplink data rate, and our devised DT-based mostly 3D map management scheme surpasses benchmark schemes in decreasing machine pose tracking uncertainty. Edge-device collaboration, AR, 3D, digital twin, deep variational inference, mannequin-based mostly reinforcement studying. Tracking the time-varying pose of each MAR machine is indispensable for MAR functions. As a result, SLAM-primarily based 3D gadget pose tracking111"Device pose tracking" can be known as "device localization" in some works. MAR applications. Despite the potential of SLAM in 3D alignment for MAR functions, limited sources hinder the widespread implementation of SLAM-based mostly 3D machine pose tracking on MAR units. Specifically, to realize accurate 3D machine pose tracking, SLAM strategies need the help of a 3D map that consists of a large number of distinguishable landmarks in the physical atmosphere. From cloud-computing-assisted tracking to the lately prevalent cellular-edge-computing-assisted monitoring, researchers have explored useful resource-environment friendly approaches for network-assisted tracking from different perspectives.

However, these analysis works have a tendency to miss the affect of community dynamics by assuming time-invariant communication resource availability or delay constraints. Treating gadget pose monitoring as a computing task, these approaches are apt to optimize networking-associated efficiency metrics resembling delay but don't seize the impression of computing job offloading and ItagPro scheduling on the efficiency of device pose tracking. To fill the hole between the aforementioned two categories of analysis works, we investigate network dynamics-aware 3D map administration for network-assisted tracking in MAR. Specifically, we consider an edge-assisted SALM structure, through which an MAR machine conducts actual-time device pose tracking regionally and uploads the captured digicam frames to an edge server. The sting server constructs and updates a 3D map using the uploaded camera frames to assist the native device pose monitoring. We optimize the performance of gadget pose monitoring in MAR by managing the 3D map, which includes importing digicam frames and updating the 3D map. There are three iTagPro key finder challenges to 3D map administration for individual MAR gadgets.

To address these challenges, we introduce a digital twin (DT)-based method to effectively cope with the dynamics of the uplink knowledge price and the device pose. DT for an MAR system to create a data mannequin that can infer the unknown dynamics of its uplink data fee. Subsequently, we suggest an synthetic intelligence (AI)-based mostly method, which utilizes the information mannequin offered by the DT to study the optimum policy for 3D map management within the presence of device pose variations. We introduce a new efficiency metric, termed pose estimation uncertainty, to point the long-term impression of 3D map management on the performance of system pose monitoring, which adapts standard system pose monitoring in MAR to network dynamics. We set up a person DT (UDT), which leverages deep variational inference to extract the latent features underlying the dynamic uplink data charge. The UDT gives these latent options to simplify 3D map administration and help the emulation of the 3D map management coverage in different network environments.

We develop an adaptive and knowledge-efficient 3D map administration algorithm featuring mannequin-primarily based reinforcement studying (MBRL). By leveraging the combination of actual information from actual 3D map management and emulated data from the UDT, the algorithm can present an adaptive 3D map administration policy in extremely dynamic network environments. The remainder of this paper is organized as follows. Section II provides an outline of related works. Section III describes the thought-about scenario and system models. Section IV presents the issue formulation and ItagPro transformation. Section V introduces our UDT, followed by the proposed MBRL algorithm based on the UDT in Section VI. Section VII presents the simulation results, and Section VIII concludes the paper. In this section, we first summarize current works on edge/cloud-assisted device pose tracking from the MAR or SLAM system design perspective. Then, we present some related works on computing job offloading and scheduling from the networking perspective. Existing studies on edge/cloud-assisted MAR applications may be categorized based on their approaches to aligning digital objects with bodily environments.

Edit

Pub: 12 Sep 2025 15:33 UTC

Views: 16