swarmsim.agent.MazeAgent#

MazeAgent has Unicycle Dynamics and can move based on sensor info.

class swarmsim.agent.MazeAgent.MazeAgentConfig(position: tuple[float, ...] | numpy.ndarray = (0, 0), angle: float | typing.Any = 0.0, name: str | typing.Any = None, controller: swarmsim.agent.control.Controller.Controller | None = None, grounded: bool = False, collides: bool | int = True, sensors: list = <factory>, team: str | None = None, seed: int | None | str = 'unspecified', agent_radius: float = 0.0, body_color: tuple[int, int, int] = (255, 255, 255), body_filled: bool = False, points: list[tuple[float, float]] | numpy.ndarray | str = <factory>, anchor_point: None | tuple[float, float] | str = None, debug: bool = False, sensing_avg: int = 1, stop_on_collision: bool = False, stop_at_goal: bool = False, catastrophic_collisions: bool = False, trace_length: int | None = 0, trace_color: tuple[int, int, int] = (255, 255, 255), track_io: bool = False, idiosyncrasies: Any = False, delay: str | int | float = 0)[source]#

Attributes:

anchor_point
controller
name
team

Methods

as_config_dict
as_dict
as_generator
asdict
attach_world_config
create
from_dict
rescale

agent_radius: float = 0.0[source]#

The radius of the agent.

Type:: float

body_color: tuple[int, int, int] = (255, 255, 255)[source]#

The body color of the agent.

Type:: tuple[int, int, int]

body_filled: bool = False[source]#

Whether the body is filled.

Type:: bool

collides: bool | int = True[source]#

Whether the agent collides with other agents.

Type:: bool | int

controller: Controller | None = None[source]#

The controller used by the agent. Zero controller by default.

Type:: AbstractController

debug: bool = False[source]#

Whether to draw the agent’s AABB in debug mode.

Type:: bool

points: list[tuple[float, float]] | ndarray | str[source]#

The points of the agent shape. If points is empty, the agent will be a circle.

Type:: list[tuple[float, float]] | np.ndarray | str

sensors: list[source]#

The sensors used by the agent. Emtpy list by default.

Type:: list[Sensor]

class swarmsim.agent.MazeAgent.MazeAgent(config: MazeAgentConfig, world, name=None, initialize=True)[source]#

Attributes:

is_poly
poly_rotated
position: Alias for pos.

Methods

`copy`()	Create a copy of this agent.
`from_config`(config, world)	Returns a new agent instance from the given config and world.
`getPosition`()	Alias for `pos`.
`getVelocity`()	Alias for `dpos`.
`get_aabb`()	Get the agent's `AABB`.
`get_sensors`()	Alias for `sensors`.
`make_aabb`()	Return the Bounding Box of the agent
`orientation_uvec`([offset])	Returns the agent's 2D orientation matrix.
`set_heading`(new_heading)	Set the agent's heading (radians).
`set_name`(new_name)	Set the name of the agent.
`set_pos_vec`(vec)	Set the `x`, `y`, and `angle` of the agent.
`setup_controller_from_config`()	Creates and adds the `AgentConfig.controller` to the agent.
`setup_sensors_from_config`()	Creates and adds sensors from the `AgentConfig.sensors` to the agent.

attach_agent_to_sensors
build_collider
debug_draw
draw
draw_direction
getFrontalPoint
get_heading
get_name
get_random_color
get_simple_poly_radius
get_x_pos
get_y_pos
handle_collisions
on_key_press
rotmat2d
rotmat2dT
rotmat3d
set_color_by_id
set_seed
set_x_pos
set_y_pos
step

DEBUG = False[source]#

SEED = -1[source]#

aabb[source]#

The agent’s cached AABB.

Type:: AABB

angle[source]#: Agent’s heading in radians. 0 is facing frame right.

attach_agent_to_sensors()[source]#

build_collider()[source]#

collides[source]#: If True, the agent should be solid.

collision_flag[source]#: Colliders should set to True if a collision was detected.

config[source]#: Agent config.

controller[source]#: The Controller for this agent.

copy()[source]#

Create a copy of this agent.

Returns:: The new agent has the same type as the class/instance it was called from.
Return type:: Agent

Almost all attributes are deep-copied from this agent. However, some attributes do not get recursively deep-copied, such as the world attribute.

The Agent._always_shallow_copy attribute is a list of agent attribute names which determines which attributes do not get deep-copied. By default, this is ['world'].

On the new agent, Attributes of this agent not in this list will be a deep-copy of those in the original agent.

Attributes of this agent in this list will share the same reference as the original agent.

Examples

>>> agent = Agent(config, world)
>>> agent.copy().world is agent.world
True
>>> copy.deepcopy(agent).world is agent.world
False

debug_draw(screen, offset)[source]#

dpos: ndarray[source]#: Change in position since last step

draw(screen, offset=((0, 0), 1.0)) → None[source]#

draw_direction(screen, offset=((0, 0), 1.0))[source]#

dt[source]#

Copy the world’s dt at agent creation.

Type:: float

dtheta[source]#: Change in heading since last step.

classmethod from_config(config, world)[source]#

Returns a new agent instance from the given config and world.

Parameters:

config (AgentConfig) – The config to create the agent from.
world (World) – This provides a back-reference to the world instance for the new agent.

Returns:

The new agent has the same type as the class/instance it was called from.

Return type:

Agent

getFrontalPoint(offset=0) → tuple[source]#

getPosition()[source]#: Alias for pos.

getVelocity()[source]#: Alias for dpos.

get_aabb() → AABB[source]#: Get the agent’s AABB.

get_heading() → float[source]#

get_name()[source]#

get_random_color()[source]#

get_sensors()[source]#: Alias for sensors.

get_simple_poly_radius()[source]#

get_x_pos() → float[source]#

get_y_pos() → float[source]#

grounded[source]#: If True, the agent should never change position or be pushed during collision resolution.

handle_collisions(world, max_attempts=10, nudge_amount=1.0, rng=None, refresh=False)[source]#

property is_poly[source]#

make_aabb() → AABB[source]#: Return the Bounding Box of the agent

name: str | None[source]#: Agent’s name.

on_key_press(event)[source]#

orientation_uvec(offset=0.0)[source]#

Returns the agent’s 2D orientation matrix.

This is a unit vector pointing in the direction of the agent’s heading plus an offset.

Equivalent to

np.array([
    math.cos(self.angle + offset),
    math.sin(self.angle + offset)
], dtype=np.float64)

Parameters:: offset (float, default=0) – The offset to add to the agent’s heading (radians).
Returns:: The 2D orientation matrix.
Return type:: numpy.ndarray (2,)

property poly_rotated[source]#

pos: ndarray[source]#: Agent’s position in the world.

property position[source]#: Alias for pos.

radius[source]#

The radius of the agent.

Type:: float

rotmat2d(offset=0)[source]#

rotmat2dT(offset=0)[source]#

rotmat3d(offset=None)[source]#

sensors: list[source]#: List of this agent’s sensors.

set_color_by_id(id)[source]#

set_heading(new_heading: float)[source]#: Set the agent’s heading (radians).

set_name(new_name: str | None | Any)[source]#

Set the name of the agent.

Parameters:: new_name (str | None | Any) – The new name of the agent. Strongly recommended to use str.

set_pos_vec(vec)[source]#

Set the x, y, and angle of the agent.

Parameters:: vec (tuple[float, float, float] | numpy.ndarray) – A vector with len() of 3 containing the x, y, angle

set_seed(seed)[source]#

set_x_pos(new_x: float)[source]#

set_y_pos(new_y: float)[source]#

setup_controller_from_config()[source]#

Creates and adds the AgentConfig.controller to the agent.

If config.controller is a dict with a type key, then it is created through the Configuration System and set as the agent’s controller. AbstractController instances copied and added to the list.

The controller is also given a back-reference to this agent via its set_agent().

Raises:: TypeError – if the element is not a dict or AbstractSensor instance.

setup_sensors_from_config()[source]#

Creates and adds sensors from the AgentConfig.sensors to the agent.

For each element of config.sensors, if the element is a dict with a type key, then it is created through the Configuration System and then added to the agent’s sensors list. AbstractSensor instances copied and added to the list.

The new sensor instances are also given a back-reference to this agent via their set_agent().

Raises:: TypeError – if the element is not a dict or AbstractSensor instance.

step(world=None, check_for_world_boundaries=None, check_for_agent_collisions=None) → None[source]#

team[source]#: The agent’s team.

world[source]#: Back-reference to the world instance.

class swarmsim.agent.MazeAgent.State[source]#

A typing.NamedTuple for recording the state of the agent.

Args:

xfloat: The agent’s x position in the world.
yfloat: The agent’s y position in the world.
anglefloat: The agent’s heading. (radians)

class swarmsim.agent.MazeAgent.SPA[source]#

A collections.namedtuple() for recording the State, Perception, Action of the agent.

Args:

stateState | tuple[float, …] | np.ndarray: The agent’s position and heading.
perceptiontuple[Any, …] | np.ndarray: The sensor readings.
actiontuple[float, …] | np.ndarray: The action to take.

Classes

`MazeAgent`(config, world[, name, initialize])
`MazeAgentConfig`(position, ...)
`SPA`(state, perception, action)
`State`(x, y, angle)

swarmsim.agent.MazeAgent#

This Page