From firstname.lastname@example.org Mon Feb 3 15:16:13 1997 Return-Path: email@example.com Date: Mon, 3 Feb 1997 09:46:56 -0500 (EST) From: firstname.lastname@example.org (Ronald Arkin) Message-Id: <199702031446.JAA19715@arkin.cc.gatech.edu> To: email@example.com Subject: Complete Rules for Find the Remote Event Cc: firstname.lastname@example.org Sender: email@example.com Precedence: bulk Status: RO
Attached below are the complete rules for the find-the-remote event. Anyone who is interested in competing in this event and has questions regarding the rules should contact Ian Horswill (firstname.lastname@example.org) for clarification.
N.B. These are not the absolute final rules. Competitors will have one month to bring up issues with the rule committee prior to their solidification.
- Ron ========
AAAI Mobile Robot Competition Event 2: Where's the remote?
This event is inspired by the need for robot assistants to perform 'fetch-it' tasks in partially known environments. Imagine a robot assistant helping a handicapped person around the home. The person might ask the robot to fetch an orange, the TV remote, a cup of coffee, and so on. While the robot may not know where all of these items are initially, over time it will learn roughly where they are kept.
The event will take place in an arena that contains tables, chairs, and shelves at varying heights. Scattered throughout the area, on the floor, the shelves, and the tables, will be 12 different objects. The robot will start the event near a human sitting in a chair (i.e.. judge) who will ask the robot to fetch three items. Once these items have been returned, the human will ask for three more items. The winner will be the robot that can find and return the most items in the allotted time.
The environment: ----------------
The environment will be divided into a kitchen-like area and a living-room-like area with a partial divider between them. The lexan walls from last year's contest will be used for the perimeter and the divider. The living-room area will include a TV and a sofa with coffee table between them. The kitchen area will include a sink, cutting board, and kitchen table. Certain items, such as the sofa and sink will necessarily be fake. Either or both areas may include other distractor items, such as chairs and trash cans, but the area will be mostly free space, so as to allow easy mobility. An example layout would be:
----------------------------------------------------- | || Sink | Cutting board| | ||-----------------------| | ||| | | |W| | | |A| | | c--||---| |L| | | o t| s | |L| K---- | | T f a| o | ||| i T | | V f b| f | t a | | e l| a | c b t | | e e| | h l r c | | |---|---| e e a a | | n --- s n | | h | -----------------------------------------------------
All the furniture listed except the trash can will be possible supporting surfaces for objects. These items will be guaranteed to be large enough to be visible on a low resolution video camera from a distance. However, they are not guaranteed to be unoccluded from the robot's initial position. Tops of supporting surfaces will be uniformly colored and will be different from the color of any object that rests on them. Pictures will be available in advance. If multiple objects are on a supporting surface, they will be well separated.
The objects: ------------
There will be twelve objects in all. Although their precise locations will not be known in advance, most objects will be constrained to lie on one of a few possible pieces of furniture. Teams are encouraged to use this context information to intelligently search the area. The objects will be:
Object Possible locations Comments ----------------------------------------
TV remote TV Black with buttons, always facing up.
Medicine bottle Coffee table or kitchen table Empty
Pillow Sofa White, large, irregular shape.
Video tape TV or coffee table A prerecorded one in a colorful box with distinctive colors
Coke can Kitchen table or coffee table Will be empty.
Coffee mug Anywhere Uniformly colored, empty.
Cereal bowl Kitchen table or sink Circular
Rubber chicken Cutting board
Fruits and veggies Sink or cutting board banana cucumber green pepper
Ketchup bottle Anywhere
Pictures of all twelve objects will be available well before the competition. Detailed color images from many different angles will be supplied for six objects. All twelve objects will be available to the teams to inspect, present to their robot, or whatever, when they arrive at the competition.
Multiple runs: --------------
- During each run of the event, the same objects will be placed on the same surface. Thus, a robot can accumulate information about the approximate location of objects from run to run.
- During each run, objects will be requested in a different order.
The robot will be started in a randomly chosen location in the rink. This location will be its "home" location to which it must return the objects. The objects will be randomly placed, but will match the location constraints given above.
Each team will rank the objects by perceived difficulty for their system. A random ordering on the objects will chosen for each team using an algorithm biased to choose easier objects first. The robot will then be given 18 minutes in which to fetch the objects in the specified order. Teams will be awarded one point for each fetched object. Teams may also opt to take the following penalties:
- Non-manipulation - 1 minute per non-manipulated object The robot need not lift the object, just drive within 1 meter of it and designate it somehow (e.g. show a camera image of the scene with an X on the object). The robot must still return home to "deliver" it.
- Virtual manipulation - 30 seconds per virtually manipulated object The team declares in advance what type of manipulator they want to simulate (e.g. a parallel-jaw gripper, dextrous hand, etc.). The robot must drive within 1 meter of the object and designate plausible contact points or areas on a video screen. There must be a consensus of the judges that the grasp is plausible. If not the object will be treated as non-manipulated (see above). The robot must still return home to "deliver" it.
- Defer - 1.5 minutes per deferred object A robot may request another object to deliver first but retain the right to deliver the deferred object at a later time.