B-Human won multiple competitions in the Middle Division of the new Humanoid Soccer League in 2026. In addition to regular matches with three or five players per team, we ventured into uncharted territory by playing demo matches with eleven humanoid robots, highlighting the tactical flexibility of our behavior and the overall robustness of our system. As B-Human formerly competed in the Standard Platform League with robots half the size and less than a third the weight, we begin by outlining the adjustments needed for this transition. The central section of this paper details our behavior architecture, drawing on concepts from RoboCup leagues that have used 11 players for much longer.
Figures & tables
Figure 1: Booster K1 robots in action during RoboCup 2026
Figure 2: Two images with marked recognized objects from the perspective of a robot during a game. Left: The violet and green rectangles show the output of the YOLOX detector. The quadrilaterals, which each have the color of the recognized jersey, show the perspective-corrected representation. The orange circle denotes the refined ball position. Right: Exemplary detection of a touchline.
Figure 3: World model of a B-Human robot (the rightmost robot, depicted together with its current field of view) during the 11 vs. 11 match at RoboCup 2026. Red and white circles with numbers denote the positions of the teammates, blue and gray circles indicate opponents and unknown robots, respectively. The currently rolling ball is drawn in orange, along with an arrow that denotes the position at which it is assumed to stop.
Figure 4: Architecture of the behavior
Figure 5: The strategy used at RoboCup 2026 for regular 3 vs. 3 and 5 vs. 5 matches. Tactics can also be invoked by selecting a derived set play (not depicted here), thus 112 may be selected. The priorities (lower = more important) are represented by circled numbers at the positions. Groups of numbers and positions connected by a dashed line mean that any of the positions may fill any of the priorities, e. g. in the 112 tactic the forward in 3 vs. 3 may be either left or right.
Figure 6: Base positions and corresponding Voronoi cells for the red team on the left, captured just after kick-off while the ball rolls from the center. Green arrows indicate the locations assigned to the red players. A cyan arrow identifies the player moving toward the ball rather than the left forward position, aiming to be the second to touch it because the HSL kick-off is indirect. The field is the one used at the RCAP Beijing Masters 2026.
Figure 7: The combined rating to evaluate passes towards teammates. The blue line and number show the value of passes, which are higher rated than dribbling or trying to score a goal directly. The violet line shows the selected pass target.
Table 1: B-Human’s game results in HSL Middle Division competitions in 2026
Physics-based humanoid control has achieved remarkable progress in enabling realistic and high-performing single-agent behaviors, yet extending these capabilities to cooperative human-object interaction (HOI) remains challenging. We present TeamHOI, a framework that enables a single decentralized policy to handle cooperative HOIs across any number of cooperating agents. Each agent operates using local observations while attending to other teammates through a Transformer-based policy network with teammate tokens, allowing scalable coordination across variable team sizes. To enforce motion realism while addressing the scarcity of cooperative HOI data, we further introduce a masked Adversarial Motion Prior (AMP) strategy that uses single-human reference motions while masking object-interacting body parts during training. The masked regions are then guided through task rewards to produce diverse and physically plausible cooperative behaviors. We evaluate TeamHOI on a challenging cooperative carrying task involving two to eight humanoid agents and varied object geometries. Finally, to promote stable carrying, we design a team-size- and shape-agnostic formation reward. TeamHOI achieves high success rates and demonstrates coherent cooperation across diverse configurations with a single policy.
Stefan Lionar, Gim Hee Lee
Garena · Sea AI Lab · National University of Singapore
Humanoid soccer is a challenging testbed for dynamic whole-body control, requiring robots to coordinate balance, locomotion, object interaction, and skill switching over long horizons. Existing humanoid sports methods often rely on task-specific multi-stage pipelines, making it difficult to jointly learn and compose multiple object-interactive skills within a single deployable policy. To address this, we present SkillX, a unified reinforcement learning framework that learns and composes multiple atomic soccer skills through a single command-conditioned policy. SkillX integrates three core designs: skill-specific adversarial motion priors, skill-specific critics, and an object-aware temporal encoder, enabling the robot to execute atomic skills and transition among them such as dribbling, trapping, and shooting. Experiments in simulation and on a real Noetix E1 humanoid demonstrate robust multi-skill execution, long-horizon skill composition, and successful sim-to-real deployment.
Many physical tasks in human environments require collaboration, from assisting a partner to jointly manipulating an object. Yet, existing humanoid benchmarks largely focus on single-humanoid skills and lack evaluation of multi-humanoid collaboration under egocentric visual observations. We introduce CoHuB (Collaborative Multi-Humanoid Benchmark), a simulation benchmark for multi-humanoid collaboration under egocentric visual observations. CoHuB provides 10 tasks, eight with two humanoids and two with three humanoids, spanning diverse collaboration patterns. We also provide synchronized demonstrations collected through a multi-operator VR teleoperation pipeline, in which each operator controls one humanoid from its egocentric view. Experiments with representative visuomotor policies reveal substantial challenges across different forms of coordinated perception and control. CoHuB provides a foundation for developing and evaluating multi-humanoid collaboration policies.
Hyunjin Park, Jebeom Chae, Minwoo Park +11
Yonsei University · Gwangju Institute of Science and Technology · University of California, Berkeley +1