MAMUT workbench local
Workspace instances · click to select
Nothing generated yet. Use the Generate tab: generated instances land in the workspace instances/ directory and appear here, ready to solve and render on the map.

Published benchmark instances live on the MAMUT-routing website; this local workbench lists the instances generated in your workspace.

1Place
Advanced
Stored OSM extracts · POI coverage
Not checked yet.

Extracts fetched before amenities mapped as building outlines were downloaded hold only the POIs mapped as points — roughly two thirds of a city's amenities. Updating re-runs the amenity query alone and leaves the road network in place.

2Customers
Sampling method
How customer locations are chosen on the road graph.
POI categories
Customers sit on real amenities of the ticked categories.
Parametric attach
Customers are sampled from road intersections, ignoring amenities.
Hybrid
Mixes both by the share below; missing POIs are filled parametrically.
Manual
You pick the exact POIs on the map. Each pick snaps to its nearest road intersection.
POI categories
0 Click POIs on the map to add them.

Click a marker to pick or unpick it; shift-click sets it as the depot. Names come from OpenStreetMap where available. Picks are snapped to the nearest road intersection, so two POIs on the same corner collapse into one customer.

Target proportion before availability fallback; missing POI customers are filled parametrically.

The pool radius is the exponential decay distance: cluster attraction halves at this distance.

Advanced
Depot position
Which road intersection becomes node 1, the depot.
Centered
Closest vertex to the mean of all vertices — short, balanced routes.
Random
A uniformly drawn vertex, so difficulty varies with the seed.
Excentered
Closest vertex to the south-west corner — long approach legs.
In manual mode you can instead click a POI and mark it as the depot.
POI attachment
How an amenity binds to the road graph, which decides how many of them can be customers at all.
Strict
The POI's nearest road node must itself be a graph intersection. Matches the published Julia pipeline exactly, and discards most of a city: an amenity whose closest node sits mid-segment is dropped even when a real intersection stands a metre further.
Snap
The POI moves to the closest intersection within the radius below, the way hand-picked POIs have always worked. Far more of the extract becomes usable; each customer records how far its amenity moved.
3Problem
Demand type
The distribution the per-customer demand is drawn from. Type 6 is spatially correlated (heavy customers cluster on one diagonal), type 7 mixes a few bulky customers into many small ones.
Route size band
The target number of customers per route. A value r is drawn uniformly from the band and sets the vehicle capacity, which in turn gives the fleet size (ceil(total demand / capacity)). It is a target, not a hard route count.
4Run
Runs the published Mamut Routing workbench pipeline locally: 3-metric CVRPLIB artifacts with shared road / geo / distance sidecars, plus VRPTW twins on request. Everything is written to your workspace.
0 instances
Selected instance
Generate an instance or pick one in the Visualize tab to display its depot and customers. Solving is optional.
MonoCost · local solve
Routes · click to toggle
Select an instance to display its customer positions. Solving adds route overlays.
Solve target
Every instance is written as three variants that differ only in arc costs, so a solution belongs to one of them and its cost is not comparable across the others.
Fastest
Arc costs are travel times in seconds.
Shortest
Arc costs are road distances in metres.
Euclidean
Arc costs are straight-line distances, ignoring the road network.
MonoCost
Minimise total routing cost.
Hierarchical
Minimise the fleet size first, then the routing cost.
Imported solutions are checked against the metric selected above before they are stored.
Validated solution runs
Solve this instance to create a validated run.

    

  

  






  

    

      
Bulk generation

      Build combinations on the left, then edit any row individually.
      
    

    

      

        
Combination builder

        

          
Cities

          

            
            
            0 selected
          

          

        

        
        

          
Demand types

          

        

        

          
Route size bands

          

        

        
        
        
        
Rows inherit the Generate form's method, categories and cluster settings; change any of
          them per row in the table. Duplicate combinations are skipped.

      

      

        

          
          
          
          
          
          
          0 instances
        

        

          

            
              

                
                Outcome
                TypeCitynDemandRoute size
                MethodSeedDepotCustomerTW method
                IntersectionsClustersDecay (m)Hybrid
                Categories
              

            
            
          

        

        

      

    

    

      
      
      
    

  







  

    POI categories
    
  

  

    
    
    
    
  

  
  

  
City fetches store every listed category. These checks only filter which POIs generation may select.







  

    About this field
    
  

  






  

    

      
Check the selection

      
    

    

    

      
      
    

  







  
  

    

      Jobs · persistent history
      
    

    
No workbench jobs yet.

    
    

      

        Log