Multi-Robot Task Allocation Using nCAR for Large CVRP Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for capacity-constrained vehicle routing problems (CVRP) do not scale well with a large number of tasks, making them impractical for applications like warehouse scenarios where hundreds to thousands of tasks need to be allocated efficiently, as they are NP-hard and do not minimize the total path traveled by robots effectively.
Innovation Solution
A processor-implemented method using the nearest neighbor-based Clustering and Routing (nCAR) approach to determine clusters of tasks that form feasible routes for vehicles, ensuring each cluster's demand is within the vehicle's capacity and minimizing the total cost, combined with a Traveling Salesman Problem (TSP) approach to optimize routes within clusters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional CVRP methods are used to allocate tasks to vehicles, then near-optimal solutions can be achieved, but the methods do not scale well with the number of tasks
Solution Approach 1:
The patent divides the large-scale CVRP problem into multiple smaller sub-problems by clustering tasks into groups. Each cluster is then solved independently using TSP, and the clusters are routed using CVRP methods. This segmentation allows the system to handle large numbers of tasks (scaling to 1000+ nodes) while maintaining solution quality, as the computational complexity is reduced from O(n) to O(n/k) where k is the number of clusters.
Solution Approach 2:
The patent performs preliminary clustering of tasks before applying CVRP methods. By pre-grouping tasks into spatially coherent clusters based on demand and capacity constraints, the system prepares the problem structure in advance to enable more efficient routing calculations. This preliminary action reduces the search space for subsequent optimization steps.
2Quantity of substance
If the number of vehicles is increased to handle more tasks, then task completion capacity is improved, but the total path length and operational cost increase
Solution Approach 1:
The patent balances the workload and path length across vehicles by creating clusters that are spatially and capacity-wise balanced. Each vehicle receives a cluster of tasks that optimizes its route, ensuring that no single vehicle is significantly overloaded or underutilized. This equipotential distribution of tasks minimizes the maximum path length any vehicle must travel while maintaining overall system capacity.
Solution Approach 2:
The patent introduces a clustering dimension to the traditional vehicle routing problem. Instead of directly routing vehicles to individual tasks, it adds an intermediate organizational layer where tasks are grouped into clusters that are then assigned to vehicles. This dimensional transformation allows for more efficient consolidation of nearby tasks and reduces redundant travel.
3Productivity
If exact CVRP methods are used to ensure optimal routing, then route efficiency is improved, but the computational time increases exponentially
Solution Approach 1:
The patent segments the routing problem into cluster-level routing and intra-cluster optimization. By solving CVRP at the cluster level rather than the individual task level, the computational burden is dramatically reduced. Each cluster contains a manageable number of tasks that can be optimized using TSP, while the overall vehicle routing is optimized using scaled-down CVRP methods.
Solution Approach 2:
The patent employs a dynamic two-stage optimization approach where the level of optimization applied depends on the problem scale. For large numbers of tasks, it uses heuristic clustering followed by localized TSP optimization, providing near-optimal solutions in polynomial time. For smaller instances, it can apply exact CVRP methods when computationally feasible, adapting the optimization intensity to the problem size.
Data Source
AI summary
Systems and methods of the present disclosure address the capacity constrained vehicle routing (CVRP) problem that may be applied to a warehouse scenario wherein multi-robot task allocation is required. Conventional methods can solve CVRP instances up to 100 nodes. In the present disclosure, a nearest-neighbor based Clustering And Routing (nCAR) approach is provided that makes the systems and methods of the present disclosure scalable wherein the number of nodes can be in the range of several hundreds to several thousands within an order wave.