Artheru：Initial bilingual draft (auto-published)

2026-05-16T14:00:23Z

Initial bilingual draft (auto-published)

新页面

<languages/>

== 概述 / Overview ==
`SimpleAgvInterface.Queue(params Func<Task>[] actions)` 是 Clumsy 车载侧的 ''任务流水线''机制。它把多个 ''异步 action''组织成可串可并的执行图，使任务作者能以声明式风格写出 "巡线 → 取货 → 升叉 → 倒车"这样的业务序列。

`SimpleAgvInterface.Queue(params Func<Task>[] actions)` is the '''task pipeline''' on the Clumsy on-vehicle side. It organises multiple async actions into a serial/parallel execution graph so mission authors can write business sequences declaratively.

实现：`D:\src\Clumsy\ClumsyCore\Interfaces\SimpleAgvInterface.cs:410`。
Implementation: `D:\src\Clumsy\ClumsyCore\Interfaces\SimpleAgvInterface.cs:410`.

== 基本用法 / Basic usage ==
<syntaxhighlight lang="csharp">
public void Fetch(double sx, double sy, double dx, double dy)
{
Queue(
async () =>
{
DriveTask.WaitDriveTask(new SteeringLineFollowing
{ srcX = currentX, srcY = currentY, dstX = sx, dstY = sy, basespeed = 600 }.Follow());
},
async () =>
{
DriveTask.WaitDriveTask(new AutoFetchGood
{ aX = sx, aY = sy, bX = dx, bY = dy, initSpeed = 500, stopDist = 80 }.Get());
},
async () =>
{
SetUpperIO("forkHeightTgt", 50);
await WaitForLowerIO("forkAtTarget", 4000);
},
async () =>
{
DriveTask.WaitDriveTask(new SteeringLineFollowingReverse
{ srcX = dx, srcY = dy, dstX = sx, dstY = sy, basespeed = 200 }.ReverseFollow());
});
}
</syntaxhighlight>

按提交顺序，4 个 action 形成 4 ''行''（rows），框架依次执行。
The 4 actions form 4 ''rows''; the framework executes them.

== 数据结构 / Data structure ==
内部维护两个队列（`SimpleAgvInterface.cs:308-352`）：
Internally there are two queues:

* '''`staticTaskLs : List<List<Task>>`''' — 静态共享队列；每行是 `List<Task>`。
* '''`taskLs : List<List<Task>>`''' — 当前实例的队列；可与 static 共享或独立。
* '''`queueInfo : List<SaiInfo>`''' — 每行的状态字符串、提交线程 ID、异常上下文。

每次 `Queue(actions)` 调用：
Each `Queue(actions)` call:

# 把 `actions[i]` ContinueWith 到 `taskLs[i]` 的最后一个 Task 后面。
# `actions[i]` 内部 `await taskLs[i-1].Last()` —— 即等上一 ''行''最新提交的 action 完成。
# 注册到 `queueInfo[i]`。

# Chain `actions[i]` after `taskLs[i].Last()` (the previous action in the same row).
# Inside `actions[i]`, `await taskLs[i-1].Last()` (the previous row's latest action).
# Register in `queueInfo[i]`.

== "超标量"实际行为 / The "super-scalar" reality ==
【发现】当前实现 ''没有''真正的跨行并发。第 i 行的 action ''必须''等第 i-1 行的 ''最新提交''完成，所以多行 Queue 调用实际上是 ''串行执行''。"超标量流水线"是设计目标，不是当前实现。
【发现】 The current implementation does NOT actually run rows in parallel. Row i awaits row i-1's most recent Task, so multi-row Queue calls execute '''sequentially'''. The "super-scalar pipeline" is design aspiration, not current behaviour.

也就是说：
In practice:

* '''单次 `Queue(a, b, c)`''' → a, b, c 串行（b 等 a 完成，c 等 b 完成）。
* '''连续 `Queue(a, b)`; `Queue(c, d)`''' → a → b → c → d 串行；c 等 b 完成（同行）；d 等 c 完成（同行）。
* '''并发要靠 action 内 `Task.WhenAll`''' —— 若你想 a + b 同时开始，自己在 a 内 `Task.WhenAll(taskA, taskB)`。

* `Queue(a, b, c)` runs a → b → c sequentially.
* Repeated `Queue` calls also serialise via row chaining.
* For concurrency '''within''' an action, use `Task.WhenAll` explicitly.

== `WaitAsync` / `Wait` ==
等待指定行完成：
Wait for a specific row to finish:

<syntaxhighlight lang="csharp">
await agv.WaitAsync(pipe: 2); // wait for row 2's latest task
// or sync
agv.Wait(pipe: 0);
</syntaxhighlight>

== `TryLock` / `Leave` ==
交管 ''锁''的两种执行路径，由 `useSimpleCallTraffic` 开关决定（`SimpleAgvInterface.cs:40`）：

Traffic locks have two paths controlled by `useSimpleCallTraffic`:

=== 同步内置 / In-process synchronous ===
当 `useSimpleCallTraffic = true`（同一进程跑调度内核 + 车载）：
* `TryLock(siteID, route_id)` 轮询 `PilotBase.latestLock == siteID`（`SimpleAgvInterface.cs:124`）
* `Leave(siteID)` 直接更新本地 `PilotBase.leaves`

=== HTTP RPC ===
跨进程时（车载与调度内核不在同一进程）：
* `TryLock` → POST `http://{host}:{port}/trylock` body `{carid, siteid}`（line 170）
* `Leave` → GET `/leave?carid={id}&siteid={siteID}`（line 260）

== 异常处理 / Error handling ==
* Action 内异常 → 写到 `queueInfo[i].exception`；该行后续 action 跳过；其它行继续。
* 调度可通过 `Flush()` 显式终结队列；之后 `Queue(...)` 抛 `_pipelineError`（line 112）。
* 任务级失败上报靠 ''任务脚本编译时插入的 try/catch'' + Hedingben toast。

* Exceptions in actions → recorded in `queueInfo[i].exception`; subsequent actions in that row skipped; other rows continue.
* `Flush()` ends the queue; later `Queue(...)` throws `_pipelineError`.

== 实战模式 / Patterns ==
=== 顺序业务动作 / Sequential business action ===
最常见。每个 action 是一段独立的 ''检查 → 行驶 → 等硬件''微操。
Most common. Each action is a self-contained "check → drive → wait hw" micro-op.

=== 锁先于动作 / Lock-then-act ===
<syntaxhighlight lang="csharp">
Queue(
async () => { if (!TryLock(targetSite, routeId)) return; /* signal failure */ },
async () => { /* the actual drive */ },
async () => Leave(prevSite)
);
</syntaxhighlight>

=== 设备 IO 等待 / Wait-for-IO ===
<syntaxhighlight lang="csharp">
Queue(
async () => { SetUpperIO("jackTarget", 80); },
async () => { await WaitForLowerIO("jackAtTarget", timeoutMs: 4000); }
);
</syntaxhighlight>

=== 错误恢复 / Error recovery ===
Action 内部 try/catch；失败时把 cart 切到 "故障"状态并 toast：

<syntaxhighlight lang="csharp">
Queue(
async () =>
{
try { await DoFragileThing(); }
catch (Exception e)
{
Hedingben.ToastText($"Action failed: {e.Message}", "AGV");
DriveStop();
throw;
}
}
);
</syntaxhighlight>

== 调度侧 AGVInterface / Fleet-side AGVInterface ==
SimpleCore 端的 `AGVInterface` (`D:\src\Simple\SimpleCore\BasicProps\AGVInterface.cs`) 提供 ''同款 Queue 机制''，用于自评估车：调度直接在自己进程里跑 Queue 来翻译 TopazScript。详见 [[Special:MyLanguage/如何基于SimpleCore核心库进行调度系统开发|如何基于SimpleCore核心库进行调度系统开发]]。

The fleet-side `AGVInterface` provides the same Queue mechanism for self-evaluating cars (scheduler runs Queue locally to translate TopazScript).

== 相关页面 / See also ==
* [[Special:MyLanguage/MovementDefinition详解|MovementDefinition详解]]
* [[Special:MyLanguage/DriveTask调度循环|DriveTask调度循环]]
* [[Special:MyLanguage/车体抽象原理|车体抽象原理]]
* [[Special:MyLanguage/如何基于SimpleCore核心库进行调度系统开发|如何基于SimpleCore核心库进行调度系统开发]]
* [[Special:MyLanguage/Clumsy-API|Clumsy-API]]

[[Category:二次开发相关说明]]
[[Category:开发手册]]

SimpleAgvInterface Queue机制 - 版本历史

Artheru：​Initial bilingual draft (auto-published)

Artheru：Initial bilingual draft (auto-published)