Multi-Robot Role Selection for Adaptive Search and Coverage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for coordinated multi-robot area search tasks often lead to suboptimal solutions and increased computational costs due to the separation of exploration and coverage tasks, which limits overall task completion efficiency.
Innovation Solution
A method for adaptively selecting roles in coordinated multi-robot search tasks using a role selection module that decouples task planning from execution, employing deep reinforcement learning to train a distributed role policy and an Actor-Critic architecture for primitive policy training, allowing robots to autonomously select roles and switch between exploration and coverage tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If exploration and coverage tasks are completely separated and executed in distinct phases, then task decomposition is simplified, but task completion efficiency decreases and computational costs increase
Solution Approach 1:
The patent segments the unified search task into two role types (exploration and coverage) that can be dynamically selected and switched between, rather than executing them as completely separate phases. This allows robots to divide their attention and actions between exploration and coverage subtasks while maintaining the ability to switch roles adaptively, thus simplifying task decomposition without sacrificing efficiency.
Solution Approach 2:
The patent implements dynamic role selection where robots can switch between exploration and coverage roles based on real-time environmental conditions and task progress. This dynamic switching mechanism allows the system to adaptively balance between simplified task decomposition and maintaining high task completion efficiency, resolving the contradiction by making the task structure flexible rather than rigid.
2Productivity
If a unified method is used to handle exploration and coverage tasks simultaneously, then resource utilization improves and efficiency increases, but computational complexity increases
Solution Approach 1:
The patent segments the complex unified task planning into distinct role selection and task execution components. By defining specific role action spaces (explore/cover) and separate perception information processing streams, the system reduces computational complexity while maintaining the benefits of unified simultaneous handling of exploration and coverage tasks.
Solution Approach 2:
The patent introduces role selection as an intermediary layer between high-level task goals and low-level execution actions. The role policy acts as a mediator that processes perception information and outputs role actions, which then guide the primitive policy for actual execution. This intermediary structure simplifies the overall computational complexity by breaking down the unified task handling into manageable stages.
3Adaptability or versatility
If task planning and task execution are coupled in a unified approach, then dynamic decision-making capabilities improve, but training complexity escalates
Solution Approach 1:
The patent segments the training process into two distinct policy training components: role policy training and primitive policy training. Each policy is trained independently with its own action space and perception information processing, which reduces training complexity while preserving dynamic decision-making capabilities through the role selection mechanism.
Solution Approach 2:
The role selection acts as an intermediary that decouples the training of high-level task planning from low-level execution. By introducing the role policy as a separate trainable component that outputs role actions, the system reduces training complexity while maintaining adaptability, as each policy layer can be trained independently rather than as a monolithic unified model.
Data Source
AI summary
The present invention discloses a method for adaptively role-selection in coordinated multi-robot search task and system thereof, comprising: defining a role action space as two discrete values: [explore, cover]; acquiring and inputting local perception information oti and joint perception information joti into a role policy, and outputting a role action ρti; wherein, the local perception information comprises an obstacle map, an explored map, a covered map, and a position map; and, the joint perception information comprises a merged explored map and a merged covered map; inputting the local perception information oti and the output role action ρti into a primitive policy, and outputting a primitive action at of robot to interact with the environment; then, a robot is controlled to execute corresponding output primitive action at according to received specific role action ρti.