flexivrobotics/flexiv_rdk

With PTP enabled, timestamps returned by Robot::states() appear approximately 4.5 ms later than the host’s PTP hardware clock read immediately afterward. Environment: Control-box software: v3.11.3 Flexiv RDK: v1.9.4 Host: Jetson AGX Orin, kernel 5.15.185-tegra LinuxPTP: 3.1.1 Direct Ethernet connection tested with GENERAL and USER_1 Robot stationary during measurement Orin’s /dev/ptp0 acts as a free-running grandmaster. No phc2sys process is running. PHC was initialized to host UTC+37 seconds before measurement; the grandmaster advertises currentUtcOffset=37, currentUtcOffsetValid=1, and ptpTimescale=1. Incoming controller Signaling messages on UDP319 were filtered because an interval request containing Announce interval 127 caused LinuxPTP to stop periodic Announce transmission. Other PTP message types remained allowed. The ptp4l configuration is: [global] time_stamping hardware network_transport UDPv4 delay_mechanism E2E domainNumber 0 utc_offset 37 free_running 1 tx_timestamp_

Hi, we are using a Rizon4 with RDK v1.9.3 and FSP v3.11.2 for visual tracking. Setup: Mode: NRT_CARTESIAN_MOTION_FORCE Command and feedback polling rate: 30 Hz Linear velocity limit: 0.01 m/s Linear acceleration limit: 0.05 m/s² Our application also applies velocity/acceleration limits to target poses before sending them. API call: robot.SendCartesianMotionForce(pose, {}, {}, 0.01, 0.1, 0.05, 0.2); Measurement: We start timing immediately before the first motion API call and stop when states().tcp_pose first shows an XY displacement of at least 1 mm from the initial position. Both timestamps use the host's monotonic clock. Across four starts: Average time to observe 1 mm displacement: 333.46 ms Range: 333.43–333.49 ms Average first API call duration: 0.059 ms The 333 ms includes acceleration, physical travel and feedback sampling; we are not treating it as pure communication latency. Questions: Is this response expected with the application-side ramp and NRT trajectory generati

Hi, I would like to clarify whether flexiv::rdk::Robot::StreamCartesianMotionForce() is officially supported on Windows without using flexiv::rdk::Scheduler. I inspected the official v1.8 Windows static library: flexiv_rdk.win_amd64.lib The library's symbol table contains a compiled definition for: flexiv::rdk::Robot::StreamCartesianMotionForce( const std::array<double, 7>& pose, const std::array<double, 6>& wrench, const std::array<double, 6>& velocity, const std::array<double, 6>& acceleration); The v1.8 robot.hpp header also declares this function without a Windows-specific guard. However, I could not find implementations of flexiv::rdk::Scheduler in the Windows library. This appears to mean that the streaming API is available, while the RDK real-time scheduler is unavailable on Windows. Could you please clarify the following? Is Robot::StreamCartesianMotionForce() officially supported on Windows in RDK v1.8? Can it be used from a user-created Windows thread or loop

Is this related to a specific version of RDK? Yes. RDK 1.9.3 with the Rizon series (software package v3.11.2). RDK 2.x requires package v3E.x and is Enlight-only, so 1.9.x is the only line available to us. What is your country or region? USA Please describe your question RDK 1.9.3 ships as a static archive only (libflexiv_rdk.x86_64-linux-gnu.a). To link it, the client application has to build and link Flexiv's entire private dependency stack. That is the wrong boundary. These are Flexiv's internal implementation choices, and the static packaging pushes all of them into our build: Fast-DDS 2.6.10, Fast-CDR 1.0.28, foonathan/memory 0.7-3, Asio RBDyn 1.9.0, SpaceVecAlg 1.2.4, jrl-cmakemodules x2, Boost 1.83.0 Eigen 3.4.0, spdlog 1.9.2, yaml-cpp 0.7.0, tinyxml2 9.0.0 That is 1.1 GB of third-party source across nine builds. Boost is 888 MB of it and is used for headers only. Our application never calls DDS, but four of those entries exist solely for Fast-DDS, and Asio is fetched only to b

Hi, We are currently using RDK v1.8. Our application is relatively simple: we execute a Plan created in Element through the RDK. We are now trying to connect a sensor to the safety input (DI 3). When the sensor is triggered, the robot enters safeguard mode. We have two questions: In the Plan execution example, busy() is used to determine whether the Plan has finished. However, when the safeguard input is triggered, the robot is paused and busy() also returns false. Is there a way to reliably distinguish between: the Plan completing normally, and the robot being paused because the safeguard input was triggered? The Rizon 4 manual mentions that the robot can automatically continue operation after the safeguard signal is cleared. How can we enable this behavior through the RDK? If possible, could you provide a minimal C++ example? Thank you for your help.