User Feedback:

• temperature, on/off state (outlet voltage), age (number of days the milk has)
• Orientation of certain objects. A computer screen might be close but you might be viewing it from behind.
• On/Off state for objects; for example a door can be locked or unlocked, a faucet might be on/off.
• In the case of the childproof system: the height difference between the child and the outlet.

User Feedback:

"I really like the interaction, but what it was not quite clear to me was what is the ”Last distance from agent” useful for. Some of the objects displayed there were not even in the perception space and still they had a distance. It did not know what I was supposed to do with the info or how to check it. Maybe a ”Current distance from the agent” would be more useful?"

User feedback:

"You can pick up the stove"

Probably the stove was configured with a really small weight which is < than the agent's maxCarryWeight

User Feedback:

• Trim the context variable for space EGOCENTRIC_CONTEXT_DATA
• regarding bugs, I couldn’t complete task 2. it is probably not a bug but the lack of providing some info that was needed. Maybe the instructions for task 2 could be a notch clearer (e.g. with examples for how the field should look when filled in for the sockets and for the pen).

User Feedback:

"When you put the coffee pot down, it looks like it’s in the stove not above it"

When placing an object onto another "surface", we should have a more advanced algorithm to compute the relative position to the given surface where the object should be placed.

User Feedback:
-Not sure how it is implemented right now but I can imagine such a simulation being a multiple step process: 1. the simulation being built by some developers. 2. After it has been developed, it is given to some ”experts” which calibrate the simulation so that it better simulate the reality of the situation it was designed for. For this they would need to easily change (possibly even at runtime) some attributes of the EgocentricContextData like: perception distance, recognition distance, examination distance, for any of the augmented objects. The framework should allow for this... I think.

Historical information has great value in a context-aware system. Implement support to save & retrieve this information through the API.

User Feedback:

Historical information about the agent: what actions has he performed so far, objects with which he has interacted with so far, etc

List of objects the agent has interacted with... or maybe last X objects

User Feedback:

• A generic USE action would help is many situations as many objects that are interacted with might have a main USE method. pickup/putdown could be specific implementations of use.

At the moment, we detect objects as being in the agent's field of vision if their BoundingBox centre is not occluded. Consider designing a more advanced algorithm!

User Feedback:

"I consider the condition for an object to enter the perception space to be too weak. A user cannot identify an object if only a ”pixel” is within his field of vision. Humans are built to recognise patterns, therefore a pattern-oriented metric could be used."

"If I am in the corner of the kitchen facing the bedroom I can see the corner of the closet but it does not appear in any other space apart from the world space. And I can use the mouse to try and interact with it and it says ”Closet, you are too far!” so maybe it should have appeared in the perception space at least."

"The living room table while standing in the bedroom. It could ”see” the cup and the laptop, but not the table."

"Standing behind the couch, it only identified the laptop, not the closet, even though it was visible. If I moved a little back, it identified the closet too."

e.g. http://www.oculusvr.com/

There should be some sort of support for this in JME already

User Feedback:

• The action space is sometimes too determined by where the cursor cross is, and not whether the object is in front of you or not. e.g., placing the cursor just on the side of the laptop makes the laptop be outside the action space. moving the cursor over the laptop makes the laptop enter the action space. the laptop should be in the action space also when the cursor is just beside it.

User Feedback:

Audio aspects of for instance the perception space.

A useful parameter would be something linked to the user’s perception of the existence of an object. For example, when you sit next to an open fire, you might not have a visual stimulus (let’s assume you have your eyes closed) that confirms that objects presence; but rather you perceive its warmth. Thus, ”perceptibility” could be defined as a measure of the subjects awareness of an object as any combination of feedback from his or her senses, in regard to said object.

User Feedback:

"The refresh is not very accurate"

User Feedback:

• the distance between all objects, so that we can have more than a visual map. Distance from edges.