-
Notifications
You must be signed in to change notification settings - Fork 0
Relativity of perception
Relativity of the perception
The perception of an agent is relative to the agent. This means that the heading of a target has to be expressed relative to the current position of the agent.
Global arena coordinates like (500, 330)
are never exposed to the agent (maybe if he has a GPS sensor ?)
Thus, indicating an POI to an agent has to be expressed as a vector from the agent to the POI (and not as the global coordinates of the agent). The heading expressed by the vector indicates the direction of the POI relative to the agent, and the magnitude of the vector indicated the distance.
This corresponds to real life; even when an agent does not know it's global coordinates, he can still walk to a tree when he sees the tree.
Limit of the perception
The agent evolves in a sphere (in 2D, a disc) of perception; POI outside it's perception range are not detectable.
The detection range varies between senses; vision has a bigger perception range that hearing, for instance.
Some senses are more subject to occlusion by obstacles than others; for instance, vision is totally blocked by occlusions, whereas sounds are only attenuated by occlusions (this attenuation might be represented by a decrease in precision of the vector to the POI, for instance) => decrease of precision in the measurement : use perlin noise to introduce smooth randomness in the measurement, as to not allow one to "mean out" the values to get a better precision (is preventing this desirable, actually ?)
Directionality of senses : differential efficiency of senses relative to the heading of the agent
Some senses are more constrained by directionality than others. Vision for instance might be restricted to a pyramid (in 2D, a triangle) originating at the agent, and pointing towards the direction he's watching; this is the Field of View (combined with the range of vision, this gives the frustum used in 3D graphics). Hearing, on the other hand, is not constrained (or negiglably so) by direction. => About directionality and field of view : https://www.youtube.com/watch?v=_ENEsV_kNx8&index=5&list=PLRqwX-V7Uu6YHt0dtyf4uiw8tKOxQLvlW
Some directional senses might be oriented by the agent without moving it's body, or with it's body pointing towards another direction (with vision for instance, this is the direction of the gaze, that may be different than the position of the body). Should this be modelled ?
Nature of the vision/hearing
- TODO: determine how to represent the nature of what is seen / heard (at least it's attractor or repulsor qualities)
- in term of ratio of attraction / repulsion ?
- in term of shape ? (seems necessary for obstacle avoidance)
- insight : might be expressed as "occluded" or "not occluded" segments on a quantized disc originating at the agent; in Super Hexagon, the agent sees occlusion through the occlusion of space in an hexagonally-sliced disc : https://www.youtube.com/watch?v=5mDjFdetU28
- in term of qualities (for exemple, the flag in a capture the flag session) ?