Table 1.
A summary of the outcomes of training and maintenance in previous studies including the I4.0 type, software, hardware, limitations, and challenges.
| Ref | I4.0 Type | Outcome | Software | Hardware | Limitations and Challenges |
|---|---|---|---|---|---|
| Chengxi et al. [12] | AR framework | Deep RL framework that is powered by AR to provide a safe and reliable HCI | Unity 3D, MRTK Toolkit3 and Vuforia Plug-In | Hololens 2 AR headset and UR5 6-Joints robot with a gripper | Limited computing power and network condition of AR devices during scene information processing, policy control inference, control command synchronization |
| Chan et al. [27] | VR framework | Improve image quality and computational efficiency of 3D GANs | PyTorch and StyleGAN2 | PC | Artifacts, lack finer details, and required knowledge of the camera pose |
| Xing et al. [23] | Markerless AR | Deep RL based on a multi-agent MADDPG to enable multi-user wireless network on an MEC server for AR application | PyTorch | Mobile devices and edge servers | The need for high-quality data to train the deep RL algorithm. |
| Hetzel et al. [24] | AR framework | First RL-based user model for mid-air and surface-aligned typing on a virtual keyboard | Python and various ML libraries | Leap motion controller and PC to run the RL algorithm | The reliance on a specific type of hand-tracking device and the collection of accurate and reliable hand-tracking data. |
| Malta et al. [18] | Automobile engines | Trained object model learns quickly and provides a prediction quickly | YOLOv5 deep neural network | AR Glasses | Trained model integration in the CMMS |
| Mourtzis et al. [31] | MR mobile application | Holographic images of the components via basic hand gestures | Cloud-based | Mobile, tablet, HMDs | The 3D component’s misalignment with respect to the actual machine |
| Mourtzis et al. [33] | MR controlling the robotic arm | Framework for controlling the robotic arm safely and remotely in almost real time | ROS | MR Microsoft HoloLens | Virtual machines and actual computers share resources, which compromises the framework’s speed |
| Alizadehsalehi et al. [15] | BIM to XR | Combines BIM and XRs to give access to all AEC stakeholders | Revit 2019/ cloud server | MR Microsoft HoloLens | The degree of stakeholder understanding of XR, software usability and willingness to pay extra for software, hardware, and training |
| Jang et al. [2] | MR touch hologram in midair | Using ultrasonic haptics with a fixed range | Ultimate immersion | VR HMD | Having the capacity to “feel” content in midair |
| Arena et al. [3] | AR-eye decides in medical applications | Makes the blood sampling easier by using portable scanners | Augmented reality markup language (ARML)/IOT | Wireless sensor network (WSN) and wireless body area network (WBAN) | Increases the AR system’s tracking precision |
| Ratcliffe et al. [34] | Remote XR | Restrictions in system development, data gathering, and participant recruiting | Unity 3D | AR-HMD (HoloLens) | Restrictions on data collection for remote XR |
| Surale et al. [32] | VR bare-hand midair mode-switching | The LEAP can accurately display hands and track their movements | Unity 3D | Standard HTC controller w | Incorrect classification of a dominant fist, palm, and pinch postures when rapidly switching mode; difficult to accurately recognize pinch actions that started manipulation |