Client-Side Latency Compensation Using Predictive Control Commands
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current communication technologies, such as TCP/IP-based connections, do not support real-time capabilities and deterministic behavior, leading to unpredictable latencies that can negatively impact the performance of real-time critical software services, especially in industrial environments with mobile devices and machines like robots and AR/VR systems.
Innovation Solution
A method for operating client devices that involves receiving predictions of sensor signals and operational commands from a server device, waiting for a predetermined time, and then operating based on the received commands or selecting the most accurate prediction if no command is received, thereby compensating for network-related latencies and ensuring real-time capability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If TCP/IP-based connections are used for cloud/edge computing, then computing power and service availability are improved, but real-time capability and deterministic behavior deteriorate due to unpredictable latencies
Solution Approach 1:
The system generates multiple predictions in advance, each comprising sensor signal hypotheses and operating command hypotheses for future time steps. These predictions are prepared beforehand so that when network latency occurs, the client device can immediately execute the most recent valid prediction without waiting for server responses, thus maintaining real-time capability while utilizing cloud computing power.
Solution Approach 2:
The system changes the parameter of time by generating predictions for multiple future time steps (t+1, t+2, ..., t+n). This allows the client device to have pre-computed commands ready for future moments, effectively compensating for network latency by having advance knowledge of what commands should be executed at future time points.
2Device complexity
If computing functions are outsourced to external servers, then local device complexity is reduced, but communication latency and network dependency increase
Solution Approach 1:
The server generates and transmits multiple predictions in advance to the client device. These predictions include operating command hypotheses that the client can execute immediately if network communication fails or experiences latency. This preliminary action reduces the need for real-time round-trip communication while maintaining the benefits of outsourced computing.
Solution Approach 2:
The system introduces an intermediary mechanism where the client device maintains a local buffer of pre-received predictions and operating command hypotheses. This intermediary layer allows the client to operate autonomously using local stored commands when network communication is delayed, effectively mediating between the outsourced server computing and local real-time execution requirements.
3Reliability
If multiple predictions are generated and stored, then latency compensation capability is improved, but memory usage and processing overhead increase
Solution Approach 1:
The system generates predictions for multiple future time steps (t+1, t+2, ..., t+n) beyond what would be minimally required. This excessive action ensures that even if several network round-trips are delayed, the client device has sufficient pre-computed predictions available. The trade-off is managed by limiting n to a practical value that balances latency compensation needs against memory constraints.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates, inter alia, to a method for operating a client device (12). If no first operating command is received from a server device (14) within a predetermined time period in response to the transmission of at least one first sensor signal, one of several predictions is selected depending on at least one first sensor signal and/or at least one currently detected sensor signal and several sensor signal hypotheses, and the client device (12) is operated with at least one operating command hypothesis of the selected prediction. The invention also relates, inter alia, to a method for operating a server device (14).