Deep Reinforcement Learning for Dependent IoV Task Offloading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current in-vehicle edge computing systems fail to consider task dependencies and energy consumption, limiting their effectiveness and efficiency in providing computational services to vehicles, especially electric vehicles, while also lacking adequate incentive compensation models.
Innovation Solution
A method utilizing deep reinforcement learning to optimize dependent task offloading in Internet of Vehicles (IoV) by constructing a vehicle system network, modeling task dependencies as a DAG, calculating delay, energy consumption, and incentive compensation, and employing a multi-queue algorithm to determine task priorities and find optimal offloading strategies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If tasks are offloaded to MEC servers in in-vehicle edge computing, then task execution delay is reduced, but edge servers have limited computation and storage capacity which cannot guarantee load balancing
Solution Approach 1:
The patent segments the task offloading problem into individual subtasks and models their dependencies using a Directed Acyclic Graph (DAG). This allows the system to handle complex task relationships by breaking them down into manageable subtask units that can be independently optimized while maintaining dependency constraints.
Solution Approach 2:
The patent employs deep reinforcement learning to create dynamic task offloading decisions that adapt to changing network conditions, vehicle states, and server capacities. The offloading strategy is not static but continuously learns and adjusts based on environmental feedback, enabling the system to handle capacity limitations dynamically.
2Power
If computational resources are concentrated at edge servers, then processing capability is improved, but energy consumption increases and incentive compensation costs rise
Solution Approach 1:
The patent applies local quality by making offloading decisions specific to each subtask based on its characteristics, dependencies, and the local state of potential destination servers. Instead of a uniform offloading strategy, the system tailors decisions to individual task requirements, optimizing the balance between processing capability and energy consumption for each specific case.
Solution Approach 2:
The patent changes the parameter of offloading decisions from static rules to dynamic values determined by deep reinforcement learning. The system learns optimal offloading parameters (which subtask goes to which server) based on varying conditions including energy states, server loads, and task characteristics, thereby optimizing the trade-off between processing power and energy consumption.
3Productivity
If current in-vehicle edge computing systems optimize individual tasks, then individual task processing is improved, but task dependencies are not considered which limits overall effectiveness
Solution Approach 1:
The patent segments tasks into subtasks and explicitly models their dependencies using a Directed Acyclic Graph (DAG) structure. This segmentation allows the system to maintain individual task optimization while simultaneously respecting dependency constraints, as the DAG clearly defines the execution order and relationships between subtasks.
Solution Approach 2:
The deep reinforcement learning framework incorporates feedback from task execution outcomes and dependency satisfaction status to continuously improve offloading decisions. The system learns from whether dependencies are met and adjusts future decisions to better balance individual task efficiency with overall task chain reliability.
Data Source
AI summary
A method of optimizing dependent task offloading in an Internet of Vehicles using deep reinforcement learning is provided, including the following steps: S1: constructing a vehicle system network; S2: constructing a task model for an application program; S3: constructing a task load model: calculating delay, energy consumption and incentive compensation for local computing, offloading to nearby vehicles and offloading to nearby RSUs three offloading methods based on the vehicle network system and the task model, respectively; S4: determining task priorities: first determining the priorities of each of subtasks according to allocation of predecessor nodes of the subtasks combining with dynamic network environment, and then scheduling according to a multi-queue algorithm; and S5: finding an optimal offloading strategy by using the deep reinforcement learning.