Mobility Competition Proposal
10 views
Skip to first unread message
Steveb
Jan 26, 2011, 2:24:46 PM1/26/11
to vma-ch...@googlegroups.com
I would like to start a discussion on how the mobility portion of the VMAC will work. The objective for this portion of the competition are to highlight teams' research and competencies while performing a relevant manufacturing/automation task. Therefore, your input is critical for the success of this event. The only constraint that I would like to set in advance of the discussion is that the USARSim UT3 version of the simulator should be used for the challenge. This means that YOUR help may be necessary to port and update components to the simulator for use in this event.
Competition Idea
I would like to start the discussion with a potential idea for the event. My idea is to have several palletizing cells located in a cluttered warehouse environment. These cells produce pallets that must be shipped to customers. For shipment to occur, the pallets must be loaded into trucks. This is accomplished by having a "kiva-like" robot that drives under the pallet and moves it to the truck.This event can start with a simple scenario and build to much more complexity. The environment could be a simple maze like world with a room that is the size and shape of a delivery truck in which the completed pallets must be placed.
Competition Rounds
Round 1: In this round, there will be multiple palletizing stations that produce pallets of goods. Teams will know the locations of the stations and the type of pallet that each station produces (e.g. station 1 produces 'palletClass1', station2 produces 'palletClass2'). A pallet will appear at the station that must be transported to the truck. Once a pallet is cleared from a station, a new one will appear after a fixed period of time. At the beginning of the run, the team will be presented with a list of pallets that must be put on the truck (e.g. 4 of palletClass1, 3 of palletClass2, 5 of palletClass3, ...). Teams will have a maximum of ? minutes to load the truck. The round ends when either time expires or all pallets have been loaded.
Round 2: In this round, the truck will make multiple trips when it unloads its goods. Therefore, the packing order of the pallets in the truck is important (pallets for the first stop must be the last loaded). Due to the production time necessary for pallets, it may be beneficial for teams to store pallets somewhere for loading onto the truck later.
Round 3: ??
Infrastructure
What infrastructure do we need? We will have a Kiva model soon (thanks to Pushkar at GT) and already have ground truth sensors (can use these to mimic an embedded tape system) and Sick sensors. Several items that I know that we need:
1) Pallet for Kiva to carry and verification that the Kiva can carry it (dock, undock, etc.).
2) Program that spawns pallets at cells to represent finished pallets
3) Automated scoring program that keeps track of time and correct/incorrect loading of pallets
Please let me know your thoughts on this idea ASAP. I am looking for additional ideas as well as volunteers to help in the creation of the missing infrastructure.
Best,
Steveb.
Competition Idea
I would like to start the discussion with a potential idea for the event. My idea is to have several palletizing cells located in a cluttered warehouse environment. These cells produce pallets that must be shipped to customers. For shipment to occur, the pallets must be loaded into trucks. This is accomplished by having a "kiva-like" robot that drives under the pallet and moves it to the truck.This event can start with a simple scenario and build to much more complexity. The environment could be a simple maze like world with a room that is the size and shape of a delivery truck in which the completed pallets must be placed.
Competition Rounds
Round 1: In this round, there will be multiple palletizing stations that produce pallets of goods. Teams will know the locations of the stations and the type of pallet that each station produces (e.g. station 1 produces 'palletClass1', station2 produces 'palletClass2'). A pallet will appear at the station that must be transported to the truck. Once a pallet is cleared from a station, a new one will appear after a fixed period of time. At the beginning of the run, the team will be presented with a list of pallets that must be put on the truck (e.g. 4 of palletClass1, 3 of palletClass2, 5 of palletClass3, ...). Teams will have a maximum of ? minutes to load the truck. The round ends when either time expires or all pallets have been loaded.
Round 2: In this round, the truck will make multiple trips when it unloads its goods. Therefore, the packing order of the pallets in the truck is important (pallets for the first stop must be the last loaded). Due to the production time necessary for pallets, it may be beneficial for teams to store pallets somewhere for loading onto the truck later.
Round 3: ??
Infrastructure
What infrastructure do we need? We will have a Kiva model soon (thanks to Pushkar at GT) and already have ground truth sensors (can use these to mimic an embedded tape system) and Sick sensors. Several items that I know that we need:
1) Pallet for Kiva to carry and verification that the Kiva can carry it (dock, undock, etc.).
2) Program that spawns pallets at cells to represent finished pallets
3) Automated scoring program that keeps track of time and correct/incorrect loading of pallets
Please let me know your thoughts on this idea ASAP. I am looking for additional ideas as well as volunteers to help in the creation of the missing infrastructure.
Best,
Steveb.
George Dimitoglou
Jan 26, 2011, 3:02:53 PM1/26/11
to VMAC
Greetings, looking forward to working with everyone ---
Here is a suggestion:
> Round 3: ??
Perhaps multiple robots loading a single truck (with the packing order
constraints from Round #2).
> *Infrastructure*
I am working with a new set of students so I doubt we can contribute
in infrastructure. For us, it is critical to have the first two items
ready as soon as possible (thanks Pushkar for the Kiva!) to get the
students working. Alternatively, for #2, if we could temporarily have
a world with a fixed panel 'spawned' (instead of waiting for a program
that spawns pallets) would be a good-enough solution to get us going.
Best,
George
Here is a suggestion:
> Round 3: ??
Perhaps multiple robots loading a single truck (with the packing order
constraints from Round #2).
> *Infrastructure*
I am working with a new set of students so I doubt we can contribute
in infrastructure. For us, it is critical to have the first two items
ready as soon as possible (thanks Pushkar for the Kiva!) to get the
students working. Alternatively, for #2, if we could temporarily have
a world with a fixed panel 'spawned' (instead of waiting for a program
that spawns pallets) would be a good-enough solution to get us going.
Best,
George
Tom Kramer
Jan 28, 2011, 3:35:45 PM1/28/11
to vma-ch...@googlegroups.com
Hi Steve -
Although my initial impression of the proposed event is positive, I am have a hard time firming up that opinion because the proposal does not contain a description of the competencies you want to test. What competencies is the "mobility portion" supposed to include? I suggest adding a section at the beginning stating what competencies are to be tested. Of course that will have to be adjusted for what is feasible to do.
For motion control:
Presumably the competencies include driving around the warehouse with or without a pallet, picking pallets up, and putting them down. Is there anything else? Does the system need to worry about being asked to move pallets that are too heavy for it? Will the size of the truck be provided apriori, or will the system have to determine the size?
For navigation:
Is accurate data on position, orientation, speed, etc. fed directly to the motion control system, or is only a limited set of information such as whether the vehicle is over a tape embedded in the floor made available? Assuming an embedded tape is used for navigation as described under "Infrastructure", there are still many alternatives. The tape might be featureless or it might have embedded position markers. The topology but not the geometry of the tape might be given, or both might be given. For geometry, there might be only qualitative data such as "the tape is composed of segments that are either very straight or are circular arcs with a radius of 1 meter". A tape known to be nice and straight could be used for determining vehicle orientation, whereas a wiggly tape could not. If something other than an embedded tape is used, the details will be different, but every alternative has details.
For map-making:
Is any kind of map-making required? For example, if the tape is given, must the system figure out where the walls of the room or aisle are, or is the tape known to be a certain (large enough) minimum distance from all walls. The "Infrastructure" section mentions Sick sensors, so presumably some use of them is intended to be required. If an apriori map is provided, is it guaranteed to have a certain level of accuracy, or might it have errors?
For obstacle detection:
Might there be obstacles in the path or is the path over the tape known to be free of obstacles? If there are obstacles, are they persistent or might they appear and disappear? Might there be other vehicles operating in the warehouse at the same time?
Under "Infrastructure":
1. Will there be only one pallet type, or will there be more than one type? Is information about the pallet type(s) provided apriori, or is the system supposed to figure it out?
2. A warehouse model will be needed. This is clearly intended; so it should be listed.
3. A model of a document or message that specifies packing order will be needed in Round 2.
Tom K.
Although my initial impression of the proposed event is positive, I am have a hard time firming up that opinion because the proposal does not contain a description of the competencies you want to test. What competencies is the "mobility portion" supposed to include? I suggest adding a section at the beginning stating what competencies are to be tested. Of course that will have to be adjusted for what is feasible to do.
For motion control:
Presumably the competencies include driving around the warehouse with or without a pallet, picking pallets up, and putting them down. Is there anything else? Does the system need to worry about being asked to move pallets that are too heavy for it? Will the size of the truck be provided apriori, or will the system have to determine the size?
For navigation:
Is accurate data on position, orientation, speed, etc. fed directly to the motion control system, or is only a limited set of information such as whether the vehicle is over a tape embedded in the floor made available? Assuming an embedded tape is used for navigation as described under "Infrastructure", there are still many alternatives. The tape might be featureless or it might have embedded position markers. The topology but not the geometry of the tape might be given, or both might be given. For geometry, there might be only qualitative data such as "the tape is composed of segments that are either very straight or are circular arcs with a radius of 1 meter". A tape known to be nice and straight could be used for determining vehicle orientation, whereas a wiggly tape could not. If something other than an embedded tape is used, the details will be different, but every alternative has details.
For map-making:
Is any kind of map-making required? For example, if the tape is given, must the system figure out where the walls of the room or aisle are, or is the tape known to be a certain (large enough) minimum distance from all walls. The "Infrastructure" section mentions Sick sensors, so presumably some use of them is intended to be required. If an apriori map is provided, is it guaranteed to have a certain level of accuracy, or might it have errors?
For obstacle detection:
Might there be obstacles in the path or is the path over the tape known to be free of obstacles? If there are obstacles, are they persistent or might they appear and disappear? Might there be other vehicles operating in the warehouse at the same time?
Under "Infrastructure":
1. Will there be only one pallet type, or will there be more than one type? Is information about the pallet type(s) provided apriori, or is the system supposed to figure it out?
2. A warehouse model will be needed. This is clearly intended; so it should be listed.
3. A model of a document or message that specifies packing order will be needed in Round 2.
Tom K.
damjan
Feb 2, 2011, 12:34:11 PM2/2/11
to VMAC
Hello everyone,
Steve, thank you for starting this discussion. In addition to the
competencies Tom has mentioned, my team (LARICS, University of Zagreb)
would like to add the following.
*Routing and scheduling*
We would like to have the efficiency of routing and scheduling
algorithms tested in the competition. In order to provide challenging
scenarios for the warehouse controller, the following four parameters
should be considered:
Number of robots - having more robots will put more emphasis on task
scheduling and conflict resolution; in our experience, 4 (or possibly
more) robots will provide some challenging scheduling situations.
Number of unloading stations - having one unloading station should be
fine, but adding more stations would make things more interesting.
Warehouse floorplan - the warehouse layout is critical in determining
the amount of both routing and conflict resolution the warehouse
controller has to perform. In previous VMA competitions, the layouts
tended to be somewhat circular in structure, which made it difficult
to assess the performance of the routing/scheduling algorithms (simply
sending the robots around in circles was usually the optimal and the
only strategy :). A rough sketch of a sample floorplan that we have in
mind can be found here: http://larics.rasip.fer.hr/dmiklic/vmac_sample_floorplan.png
We're expecting the pallets to be approximately 1m X 1m, so the
corridors on the plan are 1.5m X 1.5m, and the rooms are 25m x 25m. Of
course, all of this is still very tentative. We are currently creating
a UT3 map with this layout, our plan is to have it finished by the
beginning of next week, and we will post it on this group as soon as
it's ready.
Tasks of different priorities - finally, having tasks of different
priorities would also create interesting situations for the scheduler
to solve. For instance, pallet dispatching would be normal priority
tasks, and battery charging would be a highest-priority task. To this
end, there could be a special room with several battery-charging
stations, as shown in the sample floorplan.
As far as the infrastructure is concerned, my team could take on the
pallet modeling, for starters. Are there any design guidelines? I
couldn't find any exact specs, but the shelves that the "order fetch"
kivas are carrying seem to be approximately 1m x 1m. How would the
kiva be picking the pallets up, will the actual rotating actuator be
implemented in the model, or will we use some simplification? Of
course, we will need the robot model, as soon as it's ready.
This is more or less what we were having in mind, we'd like to hear
how the others feel about it.
Looking forward to working with everyone and having a great
competition :)
best,
damjan
> > *Competition Idea*
Steve, thank you for starting this discussion. In addition to the
competencies Tom has mentioned, my team (LARICS, University of Zagreb)
would like to add the following.
*Routing and scheduling*
We would like to have the efficiency of routing and scheduling
algorithms tested in the competition. In order to provide challenging
scenarios for the warehouse controller, the following four parameters
should be considered:
Number of robots - having more robots will put more emphasis on task
scheduling and conflict resolution; in our experience, 4 (or possibly
more) robots will provide some challenging scheduling situations.
Number of unloading stations - having one unloading station should be
fine, but adding more stations would make things more interesting.
Warehouse floorplan - the warehouse layout is critical in determining
the amount of both routing and conflict resolution the warehouse
controller has to perform. In previous VMA competitions, the layouts
tended to be somewhat circular in structure, which made it difficult
to assess the performance of the routing/scheduling algorithms (simply
sending the robots around in circles was usually the optimal and the
only strategy :). A rough sketch of a sample floorplan that we have in
mind can be found here: http://larics.rasip.fer.hr/dmiklic/vmac_sample_floorplan.png
We're expecting the pallets to be approximately 1m X 1m, so the
corridors on the plan are 1.5m X 1.5m, and the rooms are 25m x 25m. Of
course, all of this is still very tentative. We are currently creating
a UT3 map with this layout, our plan is to have it finished by the
beginning of next week, and we will post it on this group as soon as
it's ready.
Tasks of different priorities - finally, having tasks of different
priorities would also create interesting situations for the scheduler
to solve. For instance, pallet dispatching would be normal priority
tasks, and battery charging would be a highest-priority task. To this
end, there could be a special room with several battery-charging
stations, as shown in the sample floorplan.
As far as the infrastructure is concerned, my team could take on the
pallet modeling, for starters. Are there any design guidelines? I
couldn't find any exact specs, but the shelves that the "order fetch"
kivas are carrying seem to be approximately 1m x 1m. How would the
kiva be picking the pallets up, will the actual rotating actuator be
implemented in the model, or will we use some simplification? Of
course, we will need the robot model, as soon as it's ready.
This is more or less what we were having in mind, we'd like to hear
how the others feel about it.
Looking forward to working with everyone and having a great
competition :)
best,
damjan
> > I would like to start the discussion with a potential idea for the
> > event. My idea is to have several palletizing cells located in a
> > cluttered warehouse environment. These cells produce pallets that must
> > be shipped to customers. For shipment to occur, the pallets must be
> > loaded into trucks. This is accomplished by having a "kiva-like" robot
> > that drives under the pallet and moves it to the truck.This event can
> > start with a simple scenario and build to much more complexity. The
> > environment could be a simple maze like world with a room that is the
> > size and shape of a delivery truck in which the completed pallets must
> > be placed.
>
> > *Competition Rounds*
> > event. My idea is to have several palletizing cells located in a
> > cluttered warehouse environment. These cells produce pallets that must
> > be shipped to customers. For shipment to occur, the pallets must be
> > loaded into trucks. This is accomplished by having a "kiva-like" robot
> > that drives under the pallet and moves it to the truck.This event can
> > start with a simple scenario and build to much more complexity. The
> > environment could be a simple maze like world with a room that is the
> > size and shape of a delivery truck in which the completed pallets must
> > be placed.
>
> > Round 1: In this round, there will be multiple palletizing stations
> > that produce pallets of goods. Teams will know the locations of the
> > stations and the type of pallet that each station produces (e.g.
> > station 1 produces 'palletClass1', station2 produces 'palletClass2').
> > A pallet will appear at the station that must be transported to the
> > truck. Once a pallet is cleared from a station, a new one will appear
> > after a fixed period of time. At the beginning of the run, the team
> > will be presented with a list of pallets that must be put on the truck
> > (e.g. 4 of palletClass1, 3 of palletClass2, 5 of palletClass3, ...).
> > Teams will have a maximum of ? minutes to load the truck. The round
> > ends when either time expires or all pallets have been loaded.
>
> > Round 2: In this round, the truck will make multiple trips when it
> > unloads its goods. Therefore, the packing order of the pallets in the
> > truck is important (pallets for the first stop must be the last
> > loaded). Due to the production time necessary for pallets, it may be
> > beneficial for teams to store pallets somewhere for loading onto the
> > truck later.
>
> > Round 3: ??
>
> > *Infrastructure*
> > that produce pallets of goods. Teams will know the locations of the
> > stations and the type of pallet that each station produces (e.g.
> > station 1 produces 'palletClass1', station2 produces 'palletClass2').
> > A pallet will appear at the station that must be transported to the
> > truck. Once a pallet is cleared from a station, a new one will appear
> > after a fixed period of time. At the beginning of the run, the team
> > will be presented with a list of pallets that must be put on the truck
> > (e.g. 4 of palletClass1, 3 of palletClass2, 5 of palletClass3, ...).
> > Teams will have a maximum of ? minutes to load the truck. The round
> > ends when either time expires or all pallets have been loaded.
>
> > Round 2: In this round, the truck will make multiple trips when it
> > unloads its goods. Therefore, the packing order of the pallets in the
> > truck is important (pallets for the first stop must be the last
> > loaded). Due to the production time necessary for pallets, it may be
> > beneficial for teams to store pallets somewhere for loading onto the
> > truck later.
>
> > Round 3: ??
>
damjan
Feb 2, 2011, 1:05:43 PM2/2/11
to VMAC
Just noticed a typo in my previous post, the rooms are actually 5m x
5m i.e. 25m^2. Come to think about it, that might be a bit on the
small side, but as I said, these values are very tentative. Anyways,
sorry about the confusion.
damjan
5m i.e. 25m^2. Come to think about it, that might be a bit on the
small side, but as I said, these values are very tentative. Anyways,
sorry about the confusion.
damjan
damjan
Feb 8, 2011, 5:08:46 AM2/8/11
to VMAC
The initial version of our sample UT3 map is available for download
here:
http://larics.rasip.fer.hr/dmiklic/VMAC_mobility_sample_floorplan.zip
Once we have the robot and the pallet ready, we will be able to tell
which changes might be necessary.
The next thing that would come in handy is a pair of MapInfo MIF/MID
files with the floorplan layout. We are assuming these will still be
given as a priori information for the warehouse controller. Steve, how
were the MIF/MID files created in previous years? By hand, or is there
some automated tools that can do that?
best,
damjan
p.s. google is shutting down filesharing on google groups, that's why
i'm posting my files on our lab's server. steve, please let me know if
there is another preferred method for sharing files for vmac.
here:
http://larics.rasip.fer.hr/dmiklic/VMAC_mobility_sample_floorplan.zip
Once we have the robot and the pallet ready, we will be able to tell
which changes might be necessary.
The next thing that would come in handy is a pair of MapInfo MIF/MID
files with the floorplan layout. We are assuming these will still be
given as a priori information for the warehouse controller. Steve, how
were the MIF/MID files created in previous years? By hand, or is there
some automated tools that can do that?
best,
damjan
p.s. google is shutting down filesharing on google groups, that's why
i'm posting my files on our lab's server. steve, please let me know if
there is another preferred method for sharing files for vmac.
Reply all
Reply to author
Forward
0 new messages