DRL Random Access for LEO Satellite Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Low earth orbit satellite networks face significant challenges with access collisions and increased access delay times due to long communication distances and propagation delays, particularly when multiple user terminals connect to a small number of satellites, necessitating a new random access method that minimizes collisions and improves performance.
Innovation Solution
A deep reinforcement learning (DRL)-based random access method is implemented, where a terminal with a processor and memory applies a DRL algorithm to determine whether to perform an access attempt or reserve time to avoid collisions, using algorithms like Actor-Critic, DDPG, PPO, or DQN, and updates based on rewards such as communication amount, collision probability, and delay time to optimize access decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If random access technique is applied to low earth orbit satellite communication, then initial access and handover can be performed, but access performance is significantly lowered due to long communication distance and propagation delay
Solution Approach 1:
The system performs preliminary actions by predicting future satellite positions and pre-calculating optimal access parameters before the actual access attempt. The DRL algorithm learns from historical access patterns and satellite movement data to prepare access strategies in advance, reducing the impact of propagation delay during critical access moments.
Solution Approach 2:
The patent implements dynamic adaptation by continuously updating access parameters based on real-time satellite positions, terminal movements, and network conditions. The DRL algorithm dynamically adjusts backoff times, preamble selection, and satellite targeting based on changing conditions, transforming the static random access process into a dynamic optimized system.
2Adaptability or versatility
If a large number of user terminals are connected to a small number of low earth orbit satellites, then network coverage is achieved, but access performance is greatly lowered due to access collisions
Solution Approach 1:
The system applies local quality by customizing access parameters for different terminal types, service requirements, and satellite conditions. The DRL algorithm learns to apply different backoff strategies, power levels, and preamble selections based on local conditions such as terminal mobility, service priority, and satellite load, rather than using uniform random access parameters for all terminals.
Solution Approach 2:
The patent implements parameter changes by dynamically adjusting multiple access parameters including backoff time, transmit power, preamble format, and target satellite selection. The DRL algorithm optimizes these parameters based on learned patterns from network conditions, terminal characteristics, and historical access outcomes, transforming fixed parameters into adaptive variables that reduce collisions.
3Reliability
If multiple access attempts are made to overcome access collisions, then access success rate improves, but access delay time increases
Solution Approach 1:
The system implements feedback mechanisms where the DRL algorithm continuously monitors access outcomes, collision rates, and network conditions to adjust future access strategies. The algorithm receives feedback about successful accesses, collisions, and rejection reasons, and uses this information to refine its policy, learning from past experiences to minimize future delays while maintaining high success rates.
Solution Approach 2:
The patent applies preliminary action by pre-calculating optimal backoff times and access parameters before collision occurs. Rather than reacting to collisions with fixed retry protocols, the DRL algorithm predicts potential collisions based on current network conditions and terminal behavior patterns, and proactively adjusts access timing to avoid collisions before they happen, reducing the need for multiple retry attempts.
Data Source
AI summary
A random access method for a terminal with the processor and the memory to access a low earth orbit satellite network formed by multiple low earth orbit satellites (LEO SAT) includes: the stage where a Deep Reinforcement Learning (DRL) algorithm is applied for a pre-set time to decide which one between the first and the second actions should be performed at every access cycle, and to perform the random access to the low earth orbit satellite network based on the above decision while learning it; and the stage where, according to the learning result of the DRL algorithm performed for above pre-set time, it decides which of the first and the second actions should be chosen when attempting to access the low earth orbit satellite network at a new access cycle and then to perform the random access to the low earth orbit satellite network according to the above choice.