How can I change Agent.pos ?

[Nebukam]

Originally asked through this issue by @chasing2moro

I change the value pos of Agent, but It does not work. Because after I do that, Agent.pos is assigned by AgentData.position immediately. So please help me： How can I change Agent.pos ?

[Nebukam]

agent.pos updating the AgentData.position is the expected behavior; pos is used to set the current position of the agent inside the simulation.

If you're trying to set a goal location for the agent, this is something you need to implement yourself: ORCA is not a pathfinding solution and the simulation works by using the agent actual position (agent.loc) combined with a desired velocity (agent.prefVelocity).
From there you run the simulation and the position of the agent will be updated according to its own velocity & other agents inside that same simulation.

There are samples available with the package, that can be installed through the package manager; I recommend looking at Setup example (flock)

Let me know if this helps!

[chasing2moro]

After I practice Setup example (flock). It still need several frames time (≈ 0.1s) to reach the goal location. It's there any solution to reach a goal location immediately？

[Nebukam]

If you don't want to simulate the movement toward the goal then setting agent.pos as you did initially would be the way to go.

Can you provide a bit more context on what you're trying to achieve?

[chasing2moro]

I implement scenario : Teleport a unit from one postion to other position immediately, and than the unit continue to walk to enemy.

[Nebukam]

The easiest way to go about that would be something along the lines of:

1. Runs every frame:

// Given two agents, A will move toward B
agentA, agentB (enemy)

// AgentB being a moving goal, make sure AgentA prefVelocity is always pointing toward it
agentA.prefVelocity = (normalized direction from agentA to agentB) * movement speed

// Only run the simulation after that, ensuring agents have the most up-to-date 'intentions'
simulation.Run(Time.deltaTime); // Run instead of Schedule

2. Runs once/when needed, but make sure it's applied before running a new simulation step

// Teleport agentA at a desired location
agentA.pos = teleport location

In that scenario, since agentA prefVelocity is kept up-to-date each frame, it should always move toward agentB no matter its position.
Note that if you teleport an agent immediately after the simulation has run (as opposed to right before), it will still take one step of simulation to adjust agentA' orientation

Since this is multi-threaded code, as soon as a new simulation step is ran, it will take a snapshot of the current agent.pos and agent.prefVelocity to do the maths. What may be happening is that you update agent.pos while the simulation step is not done running yet, and once the step is finished it overrides the position you set manually with the simulation result.

In order to make the simulation code run synchronously (i.e without multithreading) you can call simulation.Run(Time.deltaTime) instead of simulation.Schedule(Time.deltaTime) -- this will force the simulation to run on the main thread and can be helpful for debugging behaviors related to execution order

[chasing2moro]

Thank you very much. I show my appreciation to your sincerely care about such a stranger me. You light up my path to solve the key problem.

[chasing2moro]

I finally got a solution that move to a goal postion immediately within 0.0001s and also without calculating Agent.prefVelocity-vector:

1. Invoke ORCA.TryComplete function
2. Assign goal pos to Agent.pos
3. Invoke ORCA.Schedule function. This will lead to AgentData.position assigned by Agent.pos , And then Agent.pos is assigned back by AgentData.position

        ORCA.TryComplete();

        Agent.pos = #goal pos#

        ORCA.Schedule(Time.deltaTime);