Skip to content

Latest commit

 

History

History
149 lines (76 loc) · 4.5 KB

README.md

File metadata and controls

149 lines (76 loc) · 4.5 KB
Raw

About Sensors

Track Sensor

A sensor to get track information. Covers 180 degrees of the vehicle's front. 19 returns with maximum of 200m range.

track

track_sensor.get_measurements()

Opponent Sensor

A sensor that only detects other vehicles. 36 return from 360 degrees with 200m range.

opponent opponents_sensor.get_measurements()

Road Deviation Sensor

Measures the following

  • Deviation from the center of the road in meters track_pos

  • Deviation from road orientation in radians angle

    vehicle.get_road_deviation_info()

Vehicle State Measurements

Vehicle state measurements include

  • Global Posisition
  • Global Orientation
  • Velocity
  • Acceleration
  • Angular Acceleration

Multi Agent Structure

In order to add another vehicle at desired velocity the following API call can be used.

        client.spawn_vehicle(actor= main_vehicle,distance=agent_rel_pos,initial_speed=0,set_speed=30)

  • actor : the actor that this new agent is going to be located relatively. If left empty, then track start location is taken as reference.

  • distance : relative distance to reference object. In meters.

  • initial_speed : initial speed in km/h.

  • set_speed : the target speed that the agent will try to maintain. km/h.

For convinience, an example agent addition has been added to simstarEnv.py. The parameters are

  • add_agent : Default False
  • num_agents: Maximum 7
  • agent_speeds : Target speeds of the agents. [0, 0, 20, 30, 30, 40, 50]
  • agent_locs : Relative agent locations to main vehicle .Default [100, 150, 200, 250, 300, 350, 400]

Controlling Multiple Vehicles By API

There are multiple ways of controlling a vehicle by another script using the API. Any client can be initiated and simstar observations can be created seperately. An example of this is provided in the environment. To test this feature, set the autopilot_agent flag to true.

To control the other agent, get obervations

observation_other_agent = env.get_agent_obs()

Then select the action accoirdin to this obervation

other_vehicle_action = rl_agent.action(observation_other_agent)

Finally, apply this action to other agent as

env.set_agent_action(other_vehicle_action)

Faster Simulation

You can enable faster simulation with the flag speed_up in the environment. Consider using Syncronous Mode when doing faster simulation.

Syncronous Mode

If enabled, this mode allows perfect synronization. In the provided environment, synronized_mode flag can be made True to start this mode.

Parameters in this mode

  • hz: Frequency of sampling from the environment. The environment will respond this many time in a second and will not return an observation until given number of frames are paseed in the simulator. Please note that, from version v1.5.3.B and onward, Simstar comes with a fixed frame rate of 60 fps. Therefore you should consider choosing a factor of 60 when choosin hz.

  • speed_up: This speeds up the simulation safely until -5x. Again consider hz and speed_up with the fixed FPS of the environment 60 fps. speed_up*hz should be a factpr of 60.

  • timeout. Sync Simulation blocks the simluator completely, even a new connection cannot be made. There is a default 30 seconds timeout. You can change this via

    client.set_sync_timeout(NewValue)

No Rendering Mode

In order to disable rendering, there are two options.

  1. You can Press "X" in Simstar window.

  2. You can run Simstar with the command line argument

Linux:

./SimStar.sh -no-render-simstar=1

Windows:

SimStar.exe -no-render-simstar=1
  1. You can start Simstar Witout a Window. In this mode, there is no way to see the simulation. You have to restart simualtion without the below parameter.

Linux:

./SimStar.sh -nullrhi

Windows:

SimStar.exe -nullrhi

Change default port

Default Simstar port is 8080. You can change default port with command line argument.

./SimStar.sh -api-port=8081

Change Track Type

There are 2 tracks provided as training tracks and one track provided as validation track. Change of the track can be made by passing the track name as a parameter to environment

Available tracks

  • Austria Grand Prix [Validation]

  • Dutch Grand Prix [Train]

  • Hungarian Grand Prix [Train]

Example track start:

SimstarEnv(track_name=simstar.TrackName.Austria)

hungary_track

Hungary Grand Prix track.