KlaskTron offers accessible platform for physical robot game learning
KlaskTron: An Open-Source Platform for Physical Adversarial Multi-Agent RL
RoboticsHardware ArchitectureMultiagent Systems
Summary
Robots can learn complex behaviors in computer simulations, but getting them to do the same in the real world is hard and expensive. The authors created KlaskTron, an affordable, open-source robot setup that plays a tabletop game called KLASK. This platform includes all parts and software so others can build and train robots to compete physically, matching simulated performance without extra training. Their work helps others study robot learning through physical robot competitions that anyone can reproduce.
What this means in practice
- •For robotics engineers: Use KlaskTron to develop and test multi-robot competitive strategies on a low-cost physical setup.
- •For robotics software developers: Train and validate multi-agent reinforcement learning agents with sim-to-real transfer using the KlaskTron digital twin and baseline models.
Authors
Aswin Ramachandran, Jona Schulz, Maurus Derungs, Carlo Angelini, Tobias Meier, Raffaello D'Andrea
Abstract
Progress in robot learning, particularly physical multi-agent reinforcement learning (MARL), is currently challenged by a limited availability of accessible, standardized benchmarks. While simulation-based MARL has produced remarkable emergent behaviors, from coordinated team play to complex tool use, translating these advances to physical systems remains difficult. A key barrier is infrastructural: physical platforms are often prohibitively expensive, require specialized facilities, or lack open mechanisms and designs for reproducibility. We introduce KlaskTron, an open-source, low-cost (<\$2500 USD), desktop-scale robotic testbed for adversarial MARL based on the dynamic tabletop game KLASK. The platform features dual CoreXY gantries with high-torque brushless motors, enabling the high-speed, precise manipulation required for competitive play. Crucially, we provide a complete ecosystem: hardware designs (CAD, BOM), a GPU-native digital twin in NVIDIA Isaac Lab for high-fidelity modeling and simulation, and a validated sim-to-real baseline using a learned neural actuator model. We demonstrate that this baseline enables zero-shot policy transfer, with trained agents exhibiting emergent adversarial behaviors. To ensure full reproducibility of our results, all project assets are released publicly: hardware CAD/BOM and assembly documentation at [https://idscethzurich.github.io/klask_hardware], inference software stack at [https://github.com/IDSCETHZurich/klask_software] and the simulation and MARL setup at [https://github.com/IDSCETHZurich/klask_rl].