Client and states¶
RSIClient is what RSIAPI drives. Use it directly only when you need the
raw shared variables or the state machine.
Reference¶
Bases: Enum
Connection states for RSIClient.
RSIClient(config_file: str, rsi_limits_file: Optional[str] = None, enable_auto_reconnect: bool = False, auto_reconnect_retries: int = 5, auto_reconnect_delay: float = 5.0, rsi_mode: str = 'relative', max_cartesian_rate: float = 0.0, max_joint_rate: float = 0.0, cycle_time: float = 0.004)
Main RSI API class that integrates network, config handling, and message processing.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
config_file
|
str
|
Path to RSI_EthernetConfig.xml |
required |
rsi_limits_file
|
Optional[str]
|
Optional path to .rsi.xml safety limits file |
None
|
enable_auto_reconnect
|
bool
|
Enable automatic reconnection on communication loss |
False
|
auto_reconnect_retries
|
int
|
Maximum reconnection attempts (0 = unlimited) |
5
|
auto_reconnect_delay
|
float
|
Base delay between retries in seconds |
5.0
|
rsi_mode
|
str
|
'absolute' or 'relative' — must match KRL RSI_MOVECORR() mode |
'relative'
|
max_cartesian_rate
|
float
|
Max mm/cycle for RKorr corrections (0 = disabled) |
0.0
|
max_joint_rate
|
float
|
Max degrees/cycle for AKorr corrections (0 = disabled) |
0.0
|
cycle_time
|
float
|
Expected RSI cycle time in seconds (0.004 or 0.012) |
0.004
|
Send start signal to NetworkProcess and run control loop.
Raises:
| Type | Description |
|---|---|
RSIClientNotReady
|
If client is not in appropriate state to start |
Stop the network process and the client thread safely.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stop_auto_reconnect
|
bool
|
Set False when called from the auto-reconnect path so the monitor thread survives (and is never self-joined). |
True
|
Reconnect the network process safely.
Stops the existing connection, creates fresh communication resources and a new NetworkProcess, and (by default) restarts the control loop in a background thread. E-stop state and runtime safety limits are preserved across the reconnect.
Block until the first valid packet is received from the robot.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
timeout
|
float
|
Maximum time to wait in seconds |
10.0
|
Returns:
| Type | Description |
|---|---|
bool
|
True if connected, False if timeout |
E-stop: block new writes and substitute safe corrections on the wire.
Delivered via a shared Value the network process checks every cycle, so it takes effect within one robot cycle (not the command-queue polling interval). In relative mode the wire carries zero deltas; in absolute mode the last-transmitted offset is held (zeroing would command a return-to-path motion).
Reset E-stop and resume normal corrections.
Pending user corrections are zeroed FIRST (relative mode) so motion cannot resume from stale values the moment the flag clears.
Zero all motion corrections at the source.
Writes directly to receive_variables, deliberately bypassing SafetyManager.validate — zeroing is inherently safe and must work while an E-stop is latched (validate raises during E-stop).
Set a safety limit at runtime — enforced at write time (parent) and clamped at send time (network process).
Enable/disable safety override in both enforcement layers.
Atomically publish correction dicts and bump the waypoint sequence.
Values are validated through the SafetyManager (raises on E-stop or limit violation, aborting the caller's trajectory). Returns the new sequence number to pass to wait_correction_applied().
Block until the network process has transmitted waypoint seq.
Cheap shared-memory polling (~0.5ms). Returns False on timeout (robot silent, E-stop latched, or reconnect in flight).