There are 4 stages for participation in the event.

Notice of Intent to Participate

• Notify Paul Rybski by email about your intent to participate. His email is on the home page (click the title banner at the top of the page to return home).
  
• Subscribe to the email list. This email list will be used to notify all teams about the event and updates to the rules. Additionally, teams will be able to participate in rules discussions as the rules are finalized.

Pre-Registration

All teams that seriously intend to participate need to send a pre-registration email to Paul Rybski (NOTE: this email can serve as the notice of intent to participate) . This will ensure that space is reserved for your team in terms of a work area where only registered teams are allowed. This is a secure area where equipment and computers can be safely stored overnight.

This email needs to contain the following information:

1. Team name
2. Team leader name (e.g. primary contact) with email and phone number
3. Team institution
4. Number of estimated team members to participate (can change later).
5. Whether the team will be requesting travel funding (an exact amount of request is not needed yet).

Qualification

Teams are highly encouraged to submit earlier than the listed deadline. There is no penalty for submitting multiple times.

Full Registration

Registration guarantees a team a spot in the competition as well as allowing the team to be present at the venue. Typically this just consists of paperwork but this may change from year to year.

More details on what is required by registration will be posted after the qualification deadline.

Teams that wish to participate in the competition must first qualify. No team will be allowed to register to attend and participate at the event until they have qualified.

Teams are STRONGLY recommended to submit their qualification materials earlier as there is no penalty for submitting early and resubmitting in the case of a failure to run the materials (for software league) or for insufficient description/capabilities (for the robot league).

The results of the qualification procedure will be used to determine who is eligible to compete at the competition. Travel funding will be provided to the top qualifying teams, and so doing well on the qualification results will help increase eligibility for that funding.

If we receive more qualification entries than we have space for in the competition, we will be ranking the qualification entries on a relative scale. There is no pre-set minimum performance allowed to qualify but instead the ranking is determined by how well each team's qualification materials/performance compare to the others and the top N teams (where N is to be determined) will receive travel funding.

Both leagues are required to participate in a qualification procedure. The software-only league and the robot league have different requirements.

Robot Qualification

Robot teams are required to submit a 2-page abstract providing a summary of their approach along with creating a short video of their robot that demonstrates basic navigation and obstacle avoidance capabilities. This will ensure that the robots have the basic capabilities to explore the competition environment safely.

Software-Only Qualification

Software teams are required to submit a 2-page abstract providing a summary of their approach along with a description of how their software will demonstrate its performance to the audience during the run. In addition, the software-only teams need to run their software on one of the datasets from the competitions below (choose either 2008-06-26, 2008-06-25, 2007-07-25, or 2007-07-24) and submit the output to the organizers.

Example image sets

• Dataset 2008-06-26

  These images were taken from a Nikon Coolpix 5900 still camera that was mounted on a tripod. These were the images that were given to the competitors in 2008 for the final event. The tripod was walked around the test environment and pictures were taken at random intervals.

  • Images taken from a Nikon Coolpix 5900 camera mounted on a tripod. These are the raw images. (44M tar/gzip archive of jpeg images @ 2594x1944)

    All of the objects are either on the floor or on the tables/chairs in the room. Here is the list of all objects.

• Dataset 2008-06-25

  These images were taken from a Nikon Coolpix 5900 still camera that was mounted on a tripod. These were the images that were given to the competitors in 2008 for the practice event. The tripod was walked around the test environment and pictures were taken at random intervals.

  • Images taken from a Nikon Coolpix 5900 camera mounted on a tripod. These are the raw images. (26M tar/gzip archive of jpeg images @ 2594x1944)

    All of the objects are either on the floor or on the tables/chairs in the room. Here is the list of all objects.

• Dataset 2007-07-25

  These images were taken from a Nikon Coolpix 5900 still camera that was mounted on a robot. These were the images that were given to the competitors in 2007 for the final event. The robot was driven manually through the test environment and pictures were taken at random intervals.

  • Images taken from a Nikon Coolpix 5900 camera mounted on the robot's base. These are the raw images. (49M tar/gzip archive of jpeg images @ 2594x1944)

    All of the objects are either on the floor or on the tables/chairs in the room. Here is the list of all objects.

• Dataset 2007-07-24

  These images were taken from a Nikon Coolpix 5900 still camera that was mounted on a robot. These were the images that were given to the competitors in 2007 for the practice event. The robot was driven manually through the test environment and pictures were taken at random intervals.

  • Images taken from a Nikon Coolpix 5900 camera mounted on the robot's base. These are the raw images. (48M tar/gzip archive of jpeg images @ 2594x1944)

    All of the objects are either on the floor or on the tables/chairs in the room. Here is the list of all objects.

   

• Dataset 2007-05-29

  These images were taken from a Nikon Coolpix 5900 still camera that was mounted on the front of our CMAssist robot. (See the 2007-02-07 dataset for more details about the camera and setup)

  • Images taken from a Nikon Coolpix 5900 camera mounted on the robot's base. These are the raw images. (106M tar/gzip archive of jpeg images @ 2594x1944)

    All of the objects are either on the floor or on the tables/chairs in the room. Here is the list of all objects.

• Dataset 2007-02-07

  These images were taken from a Nikon Coolpix 5900 still camera that was mounted on the front of our CMAssist robot. The libgphoto2 API under Linux was used to automatically capture these images. Because of the higher quality of these images, we will be using these for qualification purposes as well as the contest.

  Here are some images that show how the camera was mounted on the front of the robot.

  image1 image2

  Here are some photos of the environment with the objects placed within it. A low table was placed in the center of the environment and the boundaries are marked with white tape. These are just to provide a wide angle view of where all of the objects are placed.

  image1 image2 image3 image4

  • Images taken from a Nikon Coolpix 5900 mounted on the robot's base. These are the raw images. (42M tar/gzip archive of jpeg images @ 2592x1944)

    All of the objects in the image are found within the white tape boundary. Here is the list of all objects.
    
    

• Dataset 2007-01-05

  These images were taken using the top-front camera of the CMAssist robots. Images are captured at a rate of 15 fps.
  See the CMAssist web pages for more details

  • Images taken from the robot's camera at a 45 degree tilt down from the horizon (11M tar/gzip archive of jpeg images @ 640x480)
    
    
  • Undistorted images taken from the robot's camera These are the same images as the previous set but the images have been corrected for optical distortion. (11M tar/gzip archive of jpeg images @ 640x480)

  All of the objects in the image are located within the boundaries of the green carpet. There is a white line that goes around the boundary of the carpet (it's actually a robot soccer field for the four-legged league) and all of the objects in question are found within it. Here is the list of all objects.

Example of the hardest problem

To repeat: these are not the difficulty of images that will be used in this year's competition.

The first image set was taken from a mobile robot equipped with a 110 degree FOV lens (which has some optical distortions) and the other was taken with a digital camera. The data sets can be downloaded here:

• Images taken from robot's camera (36M tar/gzip archive of jpeg images @ 640x480)
  
  
• Undistorted images from robot's camera This is the same dataset as the previous one, but the images have been corrected for optical distortion. (36M tar/gzip archive of jpeg images @ 640x480)
  
  
• Images taken from a still camera (20M tar/gzip archive of jpeg images @ 2048x1536)

The following list of items can be found in the image. Some are generic classifications of objects and some are very specific.

Feedback (comments, questions, requests) about the images are welcome.