This is the issues log that is publicly viewable from the MRSD Website (if you are visiting from there, hello!)
| Item No. | Date Initiated | Date Resolved | Participants | Description | Alternative Options | Resolution | Justification |
|---|---|---|---|---|---|---|---|
| 1 | 2/4/2025 | 2/5/2025 | Tom, Kshitij, Abhishek | The interface between perception and planning subsystems needs to have commonly defined messages to avoid version mismatches and integration issues. | – Create additional repository housing commonly defined message types (Chosen)– Create and manually sync data types on both sides (likely to mismatch) | We created a Pepper.msg datatype and defined it to contain a Header, a Fruit, and a Peduncle, the latter two containing a Pose and a SolidPrimitive for shape and pose. | Subsystem leads established common understanding of what the data represents, and having one source of truth is good practice compared to separately defining message types. |
| 2 | 2/10/2025 | 4/15/2025 | Tom | Our usage of the xarm_ros repository results in high dependency on their framework and large amounts of files involved in builds | – Fully map out package dependencies and only import from xarm_ros what is strictly necessary (Chosen)– Continue working in our fork of xarm_ros repository and build additional packages around it (messy) | We are still working on migrating our code out of xarm_ros, but several packages from this repository are critical to operation and must be used. | Getting rid of unused dependencies simplified the codebase and makes it more understandable. |
| 3 | 3/20/2025 | 4/3/2025 | Abhishek | The gripper end effector was damaged, and needs to be manufactured and reassembled, which depends on sponsor knowledge | – Ask sponsor for updated CAD files, and run through production of each spare part at least once (Chosen)– Only manufacture what is broken at the moment | With help from the lab personnel, spare parts for each part of the end effector were manufactured with the lab printer and assembled. | This allows us to reduce the grey area of knowledge that we need to fetch from Dominik ahead of his departure. |
| 4 | 3/20/2025 | 3/22/2025 | Tom | There is no simple way to make sure build environments are identical between the lab PC and teammate laptops. There is a high demand for time on the shared lab PC between everyone | – Create a Dockerfile that encompasses all dependencies and have team members develop using this docker (Chosen)– Use an access schedule to ration access to the PC (restricts efficiency) | A Dockerfile was created with all dependencies for each subsystem, and used each time software is developed on team member computers. The dockerfile is updated as more packages are needed. | This allows us to parallelize our work and avoid resource constraints so that everyone can work on their own devices. The standardized dependencies in the docker allows safer development as well, as the containers can be re-created. |
1. Build Environment + Warthog:
| Item No. | Date Initiated | Date Resolved | Participants | Description | Resolution |
|---|---|---|---|---|---|
| 1. | 05.20.2025 | 06.20.2025 | Tom | Need to ensure build environments are identical between the lab PC and teammate laptops. |
| A Dockerfile was created with all dependencies for each subsystem, and used each time software is developed on team member computers. | | 2. | 08.25.2025 | 09.03.2025 | Keerthi | Need to setup OnBoard compute for mobile base with better AI inference. | Flashed Jetson Orin with Jetpack 6.2 and deployed docker containers. Verified packages are successfully installed. | | 4. | 09.29.2025 | 10.01.2025 | Abhishek | Some of the rails were not being cut in a straight orientation which was messing up intermediate measurements. | Adjusted lengths continuously to match the overall structure and made it robust | | 3. | 10.28.2025 | 10.29.2025 | Tom | Warthog computer stopped responding and added extra time to what was meant to be used for integration. | Replaced the CMOS battery and reconfigured computer, with Francisco’s help |
2. Perception :
| Item No. | Date Initiated | Date Resolved | Participants | Description | Resolution |
|---|---|---|---|---|---|
| 1. | 09.03.2025 | 09.20.2025 | Kshitij | Inefficient data augmentation and training routines. | Retrained the segmentation model for augmented data . |
| 2. | 09.10.2025 | 09.15.2025 | Keerthi | High data rate from dual realsense cameras stress out the USB bandwidth causing issues while simultaneous streaming. | Lower resolution and FPS on each camera resolves the issue allowing both the streams to be viewed simultaneously. |
| 3. | 10.06.2025 | 10.06.2025 | Keerthi | Embedding camera frames directly into the UI has synchronization issues and cause lag. | Spawning Realsense Viewer sessions within the Ui provided real-time stream. |
3. Planning & Control :
| Item No. | Date Initiated | Date Resolved | Participants | Description | Resolution |
|---|---|---|---|---|---|
| 1. | 09.01.2025 | 09.09.2025 | Abhishek, Tom | The multi-objective cost function in the planner minimizes joint movement but does not optimize task-space motion. | Developed a robust automated testing framework and implemented a multi-objective cost function. |
| 2. | 10.29.2025 | 10.29.2025 | Tom , Abhishek | Vader shared cost which worked in sim did not work on the real arms due to smaller system level errors. | Resorted to use vanilla RRT for testing, yet to be fixed |
3. Teleoperation :
| Item No. | Date Initiated | Date Resolved | Participants | Description | Resolution |
|---|---|---|---|---|---|
| 1. | 09.03.2025 | 09.09.2025 | Rohit | Calibration of gello arm was unstable leading to minor deviations in the arm’s positioning. | Made calibration sturdy for a single teleoperation arm and extensively tested calibration. |
| 2. | 09.12.2025 | 22.09.2025 | Rohit | Unable to detect teleops joint angles from dynamixel in repository due to significant code changes in simulation. | Created a separate repository fork for controlling the real arm, and the issue got resolved. |
| 3. | 10.04.2025 | 10.04.2025 | Abhishek, Rohit, Kshitij | Alignment of the XArm7 on the plate was difficult on horizontal position while installing on the Warthog. | Align the plate with the XArm7 in vertical position and slide the plate through the 8020 slot. |
| 4. | 10.28.2025 | 10.28.2025 | Rohit, Abhishek | Dual arm teleop was not working and sending the arms to a random pose. | Collision threshold was increased, input frequency was reduced and that fixed the problem |
| 5. | 10.29.2025 | 10.29.2025 | Abhishek | Xarm with cutter was not locking joints with Estop. | Adjusted payload compensation which fixed the issue |
5. End Effector :
| Item No. | Date Initiated | Date Resolved | Participants | Description | Resolution |
|---|---|---|---|---|---|
| 1. | 10.04.2025 | 10.04.2025 | Kshitij, Tom | One of the fingers fails to retract after a tight grip of a pepper | Gripper tendons were in contact with wires in the motor housing. Friction caused jamming. Check the wires within the housing |
| 2. | 10.20.2025 | 10.20.2025 | Rohit, Tom | End effector integration with teleop based control was failing | Modifying Dynamixel Driver and implementing locking mechanisms for simultaneous data communication between teleop and end effector resolved this issue. |