Control Interfaces
Rigby has two normal input routes. Both end at the same drive and steering controllers, but they accept different messages. Select one in rigby.toml; do not run both runtimes against the serial ports.
PEAK-CAN / DDT Mimic
The default peak_can route leaves the Jetson's normal stack and ros_can on the AI side. Rigby implements the vehicle/VCU side using Linux SocketCAN, normally can0, classic 11-bit CAN at 500 kbit/s.
The implementation is pinned to FS-AI API revision 33bcb61d2519ee1ac4935396d9bf465c6ebc81c1. Its exact byte layouts are in FT-Rigby's protocol contract. Do not substitute a newer DBC without reviewing the codec and tests.
| AI command ID | Purpose | Required DLC |
|---|---|---|
0x510 |
Status, handshake, mission, direction and stop request | 8 |
0x511 |
Front drive torque and RPM ceiling | 4 |
0x512 |
Rear drive torque and RPM ceiling | 4 |
0x513 |
Steering angle | 2 |
0x514 |
Brake request | 2 |
All five are required, including frames whose values are zero. Wrong-length frames do not refresh their timers. The implemented command and handshake deadlines are 100 ms; these are Rigby's test-interface settings, not a declaration of competition-rule compliance.
Feedback includes status, accepted requests, measured steering/speed, wheel counts and diagnostic/logger frames. Most feedback runs at 100 Hz, with logger status at 10 Hz. Emulated torque capability and DDT wheel geometry are protocol values, not Rigby's measured torque rating or physical wheel dimensions. Hydraulic brake feedback remains zero without pressure sensors.
From OFF to DRIVING
The implemented sequence requires fresh command frames, a completed handshake, valid actuator telemetry and the required local ASSI reports. READY lasts at least five seconds. Entry to DRIVING also checks neutral/zero requests and steering near centre.
In the normal ASSI path, the local GO signal is an eligible-window SIGUSR1, followed by an ASSI D report. An early signal is not queued for later. This description is not a launch command: the open audit findings include stop-event and ASSI-state defects.
Communication faults are intended to enter a latched emergency state. Software state changes do not give Rigby the competition car's physical emergency-braking hardware.
Drive Translation
The Jetson requests torque and an RPM ceiling, not a linear speed. The mimic takes the lower front/rear RPM ceiling, uses the pinned DDT conversion, and applies Rigby's local speed cap. Zero torque disables drive output rather than maintaining an old speed request; it allows coasting, not hydraulic braking.
Reverse remains disabled by default. Simultaneous drive/brake requests and other implausible combinations are handled by the emulated state machine. This is a software contract to test, not proof of stopping distance.
ROS Ackermann
The optional ros_ackermann route accepts ackermann_msgs/AckermannDriveStamped on /cmd over the ROS network. It does not require PEAK-CAN. The Jetson and host need a compatible ROS environment/domain and actual network connectivity.
By default, the bridge integrates the message's acceleration into a bounded speed target (use_speed_field = false). Setting use_speed_field = true selects the speed field instead. Steering arrives in radians and is converted to controller degrees. Check the selected launch parameters before publishing: the right topic is not enough if the publisher and bridge disagree about which field drives the speed target.
Command receipt must remain fresh within the configured 0.5 s window. The audit found a ROS-clock freshness problem and an unresolved active-steering-centre behaviour on authority loss. Do not describe those intended checks as fully validated fail-safe behaviour.
Steering Sign and Feedback
The external convention is positive left. Rigby's actuator convention is the opposite. With invert_steering = true, a +7.5 degree external request becomes -7.5 degrees at the steering controller, and measured feedback is converted back.
Keep the sensor calibration separate from the operational limit. The host may permit only +/-7.5 degrees while the potentiometer calibration spans +/-17.5 degrees. Actual overshoot must remain visible; clamping feedback to the request limit would hide it.
Physical Integration
Virtual CAN, simulated serial endpoints and localhost ROS tests are useful, but do not check connectors, termination, wiring, motor direction or the real Jetson link. Use the PEAK setup reference and FT-Rigby's acceptance guide, then record the physical test separately.