Autonomous Vehicle Clustering for Scalable Motion Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing centralized vehicle control methods face challenges in efficiently managing a scalable number of vehicles, particularly in motion planning, where extensive bookkeeping and lack of load balancing lead to computational inefficiencies and delays, especially when vehicles relocate across regions.
Innovation Solution
The method involves partitioning vehicles into disjoint clusters based on predefined motion constraints, using a tree-based planning algorithm, which reduces computational effort by limiting vehicle interactions across clusters and allowing for parallelization and load balancing, while coordinating resource utilization through overlapping resource subsets and prioritization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If vehicles are managed as a single centralized cluster, then comprehensive coordination is achieved, but computational complexity increases exponentially making the system intractable for large numbers of vehicles
Solution Approach 1:
The patent divides the fleet of vehicles into multiple disjoint clusters based on spatial regions and motion constraints. Each cluster is managed independently by its own planning algorithm, reducing the computational complexity from exponential (managing all vehicles together) to manageable levels (managing smaller subsets). This segmentation allows the system to scale to large numbers of vehicles while maintaining tractable computation.
Solution Approach 2:
The patent introduces a spatial dimension to the clustering strategy by dividing the operational space into regions. Vehicles are assigned to clusters based on their current spatial location and motion constraints, adding a geographic dimension to the organization structure. This spatial dimensioning allows natural partitioning of vehicles into manageable groups without arbitrary segmentation.
2Device complexity
If vehicles are partitioned into disjoint clusters, then computational tractability is improved, but coordination of shared resources across clusters becomes challenging
Solution Approach 1:
The patent introduces coordination mechanisms that act as intermediaries between disjoint clusters. These mechanisms manage shared resources (such as charging stations, parking spaces, or road segments) that are accessed by vehicles from multiple clusters. The intermediary layer handles resource allocation and conflict resolution without requiring full centralized control, thus maintaining computational tractability while enabling cross-cluster resource coordination.
Solution Approach 2:
The patent creates resource subsets that can be shared across multiple clusters, allowing the same resource to serve vehicles from different clusters under unified management rules. This universal approach to resource management enables efficient utilization of shared infrastructure while maintaining the benefits of clustered organization. Resources like corridors, waypoints, or facilities can be accessed by multiple clusters with coordinated access protocols.
3Speed
If motion planning is performed separately for each region, then planning speed is improved, but load balancing is lost causing delays when vehicles accumulate in one region
Solution Approach 1:
The patent implements dynamic cluster assignments where vehicles can be moved between clusters based on current load conditions and spatial distribution. When one region accumulates too many vehicles causing planning delays, the system dynamically reassigns vehicles to underutilized clusters or regions. This dynamic adjustment maintains planning speed by keeping individual cluster sizes manageable while achieving load balancing across the entire fleet.
Solution Approach 2:
The patent incorporates feedback mechanisms that monitor vehicle distribution across regions and planning performance metrics. Based on this feedback, the system adjusts cluster assignments and resource allocation to maintain balanced loads. When feedback indicates accumulation in one region, the system responds by redistributing vehicles, thus preventing planning delays while maintaining the speed benefits of regional partitioning.
Data Source
Figure 1~4
Figure 2
Figure 3
AI summary
There is provided a method for controlling a plurality of vehicles (v1, v2, v3, v4) sharing a set of common resources (wp1, wp2, wp3, wp4, wp5, wp6, wp7, wp8), wherein each vehicle's motion is controllable by a finite set of predefined commands (v1.a1, v1.a2, v2.a1, v2.a2). The method comprises: forming subsets ({wp1, wp2, wp3, wp4, wp5}, {wp3, wp6, wp7, wp8}) of the common resources on the basis of predefined motion constraints of individual ones of the vehicles; partitioning the vehicles into disjoint clusters ({v1, v2}, {v3, v4}) on the basis of said subsets; initiating multiple executions of a tree-based planning algorithm, each execution including an evaluation of sequences of the predefined commands to be fed to the vehicles in a single cluster and a selection of a preferred sequence of the commands; and feeding said preferred sequence of commands to the vehicles in each cluster.