as most of you should already know by now, we use to publish all our work each year during or after each RoboCup since we first participated in 2011.And we continue with this attitude, because we belief in the importance of sharing knowledge to advance our league and the RoboCup.
Below, you will find a list of our publications and a list of the most significant changes. The documentation or description is included within the individual packages.
Modified the curved front shape of the straight kicking plunger. This
significantly reduced the spread of kicked balls when they are not
perfectly at the center in front of the robot.- Enlarged the
infrared barrier PCBs on the sides of the dribbling device to prevent
the silicon rubbers of the linear damping mechanism to fall out
Electronics– Usage of a new wireless module: Semtech SX1280- Schematics and PCB layout of a tiny 34x18mm module with a single SX1280 module. Pin-compatible with most nRF24L01+ breakout boards on the market (SX1280.pdf).- Schematics and PCB layout of a completely new base station using one SX1280 module and the SKY66112 front-end module for enhanced output power and antenna switching. Features network connectivity and a 5″ touchscreen with 800×480 pixels.Release: https://tigers-mannheim.de/download/release2018/Electronics-2018.zip
Firmware– Support for the SX1280 and SKY66112 modules (in base station and mainboard code).- Enhanced wireless transfer protocol to make best use of the new modules.- Adaption of the base station code to the new hardware (incl. completely new user interface).- Base station touchscreen supports manual control of a single robot for demonstration purposes.- Base station can be configured via touchscreen.- New feature: Asynchronous Bang-Bang trajectories. They are intentionally not synchronized on the X and Y trajectory to reach a specific line/direction faster.- Major update of robot controller and sensor fusion. The robot now uses a state space controller with acceleration output and model-based friction compensation when a position source (e.g. SSL Vision) is available. If not position source is available, local velocity output is used with a velocity-error correction for encoder readings. All models are derived from empirical data.Release: https://tigers-mannheim.de/download/release2018/Firmware-2018.zip
Software (TIGERs AutoRef)– Detect more rule violations- Updated to new rules- Significantly improved ball speed detection and chip kick detection (see Vision Filter)Code: https://gitlab.tigers-mannheim.de/open-source/AutoReferee
Software (AI)– Enhanced the integrated AutoRef for full automatic simulation of two teams playing against each other- Offensive Action Trees for self learning offensive behaviors (see ETDP)- Multiple new supportive behaviors for supporting the attacker- Improved the selection of the best suited attacker that should handle the ball- Improved the skills that handle the ball, like kicking, receiving, redirecting, approaching and catching- Integrated new asynchronous bang bang trajectories into multiple skills- More robust path planning with ahead-of-time emergency brakeRelease: https://tigers-mannheim.de/download/release2018/Software-2018.tar.gz
Software (Vision Filter)– New chip kick model: We are now using fixed restitution coefficients for all hops in Z direction, for the first hop in XY direction, and for all further hops in XY direction (3 parameters in total).- Online straight kick and chip kick estimation has been updated to use a direct simplex-based optimization method.- Chip kick estimation can now optimize a complete trajectory, not only two hops.- An additional optimizer can estimate all model parameters for a straight and chip kick with an evolutionary algorithm (CMAES). Model parameters can now be fully calculated within Sumatra, no further external software required.- The Vision Filter is completey integrated into our official AutoRef software.- Kick information from our own robots is now used as a-priori knowledge for estimation.- The viewport architect, which aligns the viewports of the different cameras and tries to minimize overlap, now supports different modes of how the viewports are calculated. Based on field size, camera positions, camera projections, or dynamically on detected robot positions.- Correction of inbound SSL Vision timestamps. The average rate of each camera is estimated and the timestamps are corrected for each camera, mitigating network and processing delays. This increases filter accuracy for ball and robots.Release: https://tigers-mannheim.de/download/release2018/Software-2018.tar.gz
Software (status-board)– A new application for visualizing referee information during a game including new game events that are announced by the autoRefs.- Additionally, a live view of the field.Code: https://github.com/RoboCup-SSL/ssl-status-board-clientApplication is available here (and will so during the RoboCup): https://tigers-mannheim.de/status-board/
Looking forward to seeing you all in Montreal 🙂
Dear SSL community,
As you may already know, we have open-sourced our electronics in January. Besides that, we would also like to share our mechanical design and vision process software. We hope these can help the SSL community. Detailed design specification and software guide can be found in the corresponding github repository listed below.
Mechanics (Include 3D Design Files)
The mechanical design emphasizes on crash protection, maintainability and simplicity. The dribbler design is thoroughly explained.
Electronics (Include Schematics, PCBs and Related Firmware)
The design files are up-to-date. We also present the transition of design from 2012 to 2019 which may provide other teams a practical way to progress.
Software (mAn – Vision Process and GUI)
mAn is the main GUI of ZJUNlict. It also processes the vision and referee information (protobuf messages).
Looking forward to see you all in Sydney!
Dear SSL community,
This year RoboJackets has started outsourcing our mechanical designs. Additionally, we are continuing our tradition of open sourcing our software, firmware, and PCB designs through GitHub. This year we have added a new project: A fully open source microcontroller board based on the Cortex M7 with an emphasis on being small enough to easily fit within small size league robots.
GUI and High Level Control: https://github.com/RoboJackets/robocup-software
Microcontroller HAL: https://github.com/RoboJackets/mtrain-firmware