Scoresheet states to approach a human's hand, so it is assumed the human will first hand the item close to the robot and then we can ask for them to hand it over
Manipulation & Mechanics & Electronics
Detect an object in the robot's gripper. 5x100 points
It will be considered a Deus Ex Machina if the robot tells the human when to leave the object in the gripper, such as counting down to 0. We have to detect the object in the robot and then warn the used that we will pick it up. We lose 75 points if the robot has to use Deus Ex Machina.
We need a working sensor to detect the object in the gripper.
Section 2: Conversation & Place
HRI
Conversational node -> 5x50 points
Receives the robot's knowledge of the environment as in GPSR
Has to initiate conversation, and get their attention.
Has to continue conversation until an unambiguous placement location is obtained -> a place location + description e.g. " the ", " the ", " the "
Has to be able to ask for clarification if the location is not clear
Vision
Has to infer visual queues as to which position the object should be placed in
If a human is pointing left, right, up.... it may be able to provide this extra information to the conversational node, connected with the estimated locations the person may be aiming at. E.g. if the robot is aiming right and this is the direction of a table and the kitchen, we can assume the person is pointing at either of them. Should this connection be addressed here or in HRI? Either way, to accomplish this a connection with the map and its semantic information is needed.
Manipulation
A list of instructions that can be added to the place task:
Place on
Place to the left of
Place to the right of
Place in front of
Place behind
Place on top of
...
Execute place tasks with said added instructions -> 5x100 points
Integration
We have to connect the instruction to real world objects, either by searching for them or connecting to the map's locations.
If instruction is " the ", we have to find the apple in the map and then place the object next to it.
If instruction is " the ", we may already know which table that is from mapped locations.
If instruction is " the ", shelf is expected to be a mapped location too.
It is necessary to address additional ambiguities that may show up on execution. For example, we can leave the conversation with the instruction to place the object " the in the ", but the robot may encounter two green bins on this place. We should be capable to ask for clarification.