Learning agile soccer skills for a bipedal robot with deep reinforcement learning.

We investigated whether deep reinforcement learning (deep RL) is able to synthesize sophisticated and safe movement skills for a low-cost, miniature humanoid robot that can be composed into complex behavioral strategies. We used deep RL to train a humanoid robot to play a simplified one-versus-one soccer game. The resulting agent exhibits robust and dynamic movement skills, such as rapid fall recovery, walking, turning, and kicking, and it transitions between them in a smooth and efficient manner. It also learned to anticipate ball movements and block opponent shots. The agent's tactical behavior adapts to specific game contexts in a way that would be impractical to manually design. Our agent was trained in simulation and transferred to real robots zero-shot. A combination of sufficiently high-frequency control, targeted dynamics randomization, and perturbations during training enabled good-quality transfer. In experiments, the agent walked 181% faster, turned 302% faster, took 63% less time to get up, and kicked a ball 34% faster than a scripted baseline.

TacticAI: an AI assistant for football tactics.

Identifying key patterns of tactics implemented by rival teams, and developing effective responses, lies at the heart of modern football. However, doing so algorithmically remains an open research challenge. To address this unmet need, we propose TacticAI, an AI football tactics assistant developed and evaluated in close collaboration with domain experts from Liverpool FC. We focus on analysing corner kicks, as they offer coaches the most direct opportunities for interventions and improvements. TacticAI incorporates both a predictive and a generative component, allowing the coaches to effectively sample and explore alternative player setups for each corner kick routine and to select those with the highest predicted likelihood of success. We validate TacticAI on a number of relevant benchmark tasks: predicting receivers and shot attempts and recommending player position adjustments. The utility of TacticAI is validated by a qualitative study conducted with football domain experts at Liverpool FC. We show that TacticAI's model suggestions are not only indistinguishable from real tactics, but also favoured over existing tactics 90% of the time, and that TacticAI offers an effective corner kick retrieval system. TacticAI achieves these results despite the limited availability of gold-standard data, achieving data efficiency through geometric deep learning.

From motor control to team play in simulated humanoid football.

Learning to combine control at the level of joint torques with longer-term goal-directed behavior is a long-standing challenge for physically embodied artificial agents. Intelligent behavior in the physical world unfolds across multiple spatial and temporal scales: Although movements are ultimately executed at the level of instantaneous muscle tensions or joint torques, they must be selected to serve goals that are defined on much longer time scales and that often involve complex interactions with the environment and other agents. Recent research has demonstrated the potential of learning-based approaches applied to the respective problems of complex movement, long-term planning, and multiagent coordination. However, their integration traditionally required the design and optimization of independent subsystems and remains challenging. In this work, we tackled the integration of motor control and long-horizon decision-making in the context of simulated humanoid football, which requires agile motor control and multiagent coordination. We optimized teams of agents to play simulated football via reinforcement learning, constraining the solution space to that of plausible movements learned using human motion capture data. They were trained to maximize several environment rewards and to imitate pretrained football-specific skills if doing so led to improved performance. The result is a team of coordinated humanoid football players that exhibit complex behavior at different scales, quantified by a range of analysis and statistics, including those used in real-world sport analytics. Our work constitutes a complete demonstration of learned integrated decision-making at multiple scales in a multiagent setting.

Multiagent off-screen behavior prediction in football.

In multiagent worlds, several decision-making individuals interact while adhering to the dynamics constraints imposed by the environment. These interactions, combined with the potential stochasticity of the agents' dynamic behaviors, make such systems complex and interesting to study from a decision-making perspective. Significant research has been conducted on learning models for forward-direction estimation of agent behaviors, for example, pedestrian predictions used for collision-avoidance in self-driving cars. In many settings, only sporadic observations of agents may be available in a given trajectory sequence. In football, subsets of players may come in and out of view of broadcast video footage, while unobserved players continue to interact off-screen. In this paper, we study the problem of multiagent time-series imputation in the context of human football play, where available past and future observations of subsets of agents are used to estimate missing observations for other agents. Our approach, called the Graph Imputer, uses past and future information in combination with graph networks and variational autoencoders to enable learning of a distribution of imputed trajectories. We demonstrate our approach on multiagent settings involving players that are partially-observable, using the Graph Imputer to predict the behaviors of off-screen players. To quantitatively evaluate the approach, we conduct experiments on football matches with ground truth trajectory data, using a camera module to simulate the off-screen player state estimation setting. We subsequently use our approach for downstream football analytics under partial observability using the well-established framework of pitch control, which traditionally relies on fully observed data. We illustrate that our method outperforms several state-of-the-art approaches, including those hand-crafted for football, across all considered metrics.

Learning a generative model of images by factoring appearance and shape.

Computer vision has grown tremendously in the past two decades. Despite all efforts, existing attempts at matching parts of the human visual system's extraordinary ability to understand visual scenes lack either scope or power. By combining the advantages of general low-level generative models and powerful layer-based and hierarchical models, this work aims at being a first step toward richer, more flexible models of images. After comparing various types of restricted Boltzmann machines (RBMs) able to model continuous-valued data, we introduce our basic model, the masked RBM, which explicitly models occlusion boundaries in image patches by factoring the appearance of any patch region from its shape. We then propose a generative model of larger images using a field of such RBMs. Finally, we discuss how masked RBMs could be stacked to form a deep model able to generate more complicated structures and suitable for various tasks such as segmentation or object recognition.

Direction opponency, not quadrature, is key to the 1/4 cycle preference for apparent motion in the motion energy model.

Sensitivity to visual motion is a fundamental property of neurons in the visual cortex and has received wide attention in terms of mathematical models. A key feature of many popular models for cortical motion sensors is the use of pairs of functions that are related by a 90 degrees phase shift. This phase relationship, known as quadrature, is the hallmark of the motion energy model and played an important role in the development of a class of model dubbed elaborated Reichardt detectors. For decades, the literature has supported a link between quadrature and the observation that motion detectors and human observers often prefer a 1/4 cycle displacement of an apparent motion stimulus that consists of a pair of sinusoidal gratings. We show that there is essentially no link between quadrature and this preference. Quadrature is neither necessary nor sufficient for a motion sensor to prefer 1/4 cycle displacement, and motion energy is not maximized for a 1/4 cycle step. Other properties of motion sensors are the key: the opponent subtraction of two oppositely tuned stages that individually have sinusoidal displacement tuning curves. Thus, psychophysical and neurophysiological data revealing a preference at or near 1/4 cycle displacement do not offer specific support for common quadrature or energy-based motion models. Instead, they point to a broader class of model.