Authoring high-quality character animation is essential in multimedia production, especially for pre-rendered formats such as films and TV series. These workflows are often iterative, requiring frequent adjustments and immediate visual feedback, and typically involve editing existing motion, such as mocap data. We present Neural Motion Path (NMP), a deep learning-based system designed to support this authoring process by enabling full-body motion editing through joint-level motion path manipulation. While joint rotations are essential for expressive human motion, explicitly specifying them imposes a significant burden on users; NMP addresses this challenge by enabling intuitive position-only motion path editing while implicitly inferring plausible rotation trajectories. To address the inherent tension between motion plausibility and precise constraint satisfaction, NMP explicitly decouples motion synthesis from constraint enforcement within an autoregressive framework. NMP generates realistic, context-aware motion without fine-tuning every path detail, making it accessible to novices while supporting the demands of detailed animation authoring. It combines a motion generator with a novel RotationNet for inferring joint rotation, and a constraint imposer that enforces end-effector constraints via an Explicit-Weight Sparse Expert Model (EW-SEM). The system supports terrain adaptation and authoring operations like Concatenate, Insert, and Mix. Implemented as a Blender add-on, NMP supports real-time playback and interactive workflows. A user study with novice users shows that NMP improves satisfaction, efficiency, and perceived motion quality compared to the conventional layered keyframing approach, highlighting its potential as an accessible and effective authoring tool.

We present TouchWalker, a real-time system for controlling fullbody avatar locomotion using finger-walking gestures on a touchscreen. The system comprises two main components: TouchWalker-MotionNet, a neural motion generator that synthesizes full-body avatar motion on a per-frame basis from temporally sparse twofinger input, and TouchWalker-UI, a compact touch interface that interprets user touch input to avatar-relative foot positions. Unlike prior systems that rely on symbolic gesture triggers or predefined motion sequences, TouchWalker uses its neural component to generate continuous, context-aware full-body motion on a per-frame basis-including airborne phases such as running, even without input during mid-air steps-enabling more expressive and immediate interaction. To ensure accurate alignment between finger contacts and avatar motion, it employs a MoE-GRU architecture with a dedicated foot-alignment loss. We evaluate TouchWalker in a user study comparing it to a virtual joystick baseline with predefined motion across diverse locomotion tasks. Results show that TouchWalker improves users' sense of embodiment, enjoyment, and immersion.

We propose FreeMusco, a motion-free framework that jointly learns latent representations and control policies for musculoskeletal characters. By leveraging the musculoskeletal model as a strong prior, our method enables energy-aware and morphology-adaptive locomotion to emerge without motion data. The framework generalizes across human, non-human, and synthetic morphologies, where distinct energy-efficient strategies naturally appear—for example, quadrupedal gaits in Chimanoid versus bipedal gaits in Humanoid. The latent space and corresponding control policy are constructed from scratch, without demonstration, and enable downstream tasks such as goal navigation and path following—representing, to our knowledge, the first motion-free method to provide such capabilities. FreeMusco learns diverse and physically plausible locomotion behaviors through model-based reinforcement learning, guided by the locomotion objective that combines control, balancing, and biomechanical terms. To better capture the periodic structure of natural gait, we introduce the temporally averaged loss formulation, which compares simulated and target states over a time window rather than on a per-frame basis. We further encourage behavioral diversity by randomizing target poses and energy levels during training, enabling locomotion to be flexibly modulated in both form and intensity at runtime. Together, these results demonstrate that versatile and adaptive locomotion control can emerge without motion capture, offering a new direction for simulating movement in characters where data collection is impractical or impossible.

We propose PhysicsFC, a method for controlling physically simulated football player characters to perform a variety of football skills-such as dribbling, trapping, moving, and kicking-based on user input, while seamlessly transitioning between these skills. Our skill-specific policies, which generate latent variables for each football skill, are trained using an existing physics-based motion embedding model that serves as a foundation for reproducing football motions. Key features include a tailored reward design for the Dribble policy, a two-phase reward structure combined with projectile dynamics-based initialization for the Trap policy, and a Data-Embedded Goal-Conditioned Latent Guidance (DEGCL) method for the Move policy. Using the trained skill policies, the proposed football player finite state machine (PhysicsFC FSM) allows users to interactively control the character. To ensure smooth and agile transitions between skill policies, as defined in the FSM, we introduce the Skill Transition-Based Initialization (STI), which is applied during the training of each skill policy. We develop several interactive scenarios to showcase PhysicsFC's effectiveness, including competitive trapping and dribbling, give-and-go plays, and 11v11 football games, where multiple PhysicsFC agents produce natural and controllable physics-based football player behaviors. Quantitative evaluations further validate the performance of individual skill policies and the transitions between them, using the presented metrics and experimental designs.

Since the introduction of DeepMimic [Peng et al. 2018a], subsequent research has focused on expanding the repertoire of simulated motions across various scenarios. In this study, we propose an alternative approach for this goal, a deep reinforcement learning method based on the simulation of a single-rigid-body character. Using the centroidal dynamics model (CDM) to express the full-body character as a single rigid body (SRB) and training a policy to track a reference motion, we can obtain a policy that is capable of adapting to various unobserved environmental changes and controller transitions without requiring any additional learning. Due to the reduced dimension of state and action space, the learning process is sample-efficient. The final full-body motion is kinematically generated in a physically plausible way, based on the state of the simulated SRB character. The SRB simulation is formulated as a quadratic programming (QP) problem, and the policy outputs an action that allows the SRB character to follow the reference motion. We demonstrate that our policy, efficiently trained within 30 minutes on an ultraportable laptop, has the ability to cope with environments that have not been experienced during learning, such as running on uneven terrain or pushing a box, and transitions between learned policies, without any additional learning.

We present a flexible and efficient approach for generating multilegged locomotion. Our model-predictive control (MPC) system efficiently generates terrain-adaptive motions, as computed using a three-level planning approach. This leverages two commonly-used simplified dynamics models, an inverted pendulum on a cart model (IPC) and a centroidal dynamics model (CDM). Taken together, these ensure efficient computation and physical fidelity of the resulting motion. The final full-body motion is generated using a novel momentum-mapped inverse kinematics solver and is responsive to external pushes by using CDM forward dynamics. For additional efficiency and robustness, we then learn a predictive model that then replaces two of the intermediate steps. We demonstrate the rich capabilities of the method by applying it to monopeds, bipeds, and quadrupeds, and showing that it can generate a very broad range of motions at interactive rates, including banked variable-terrain walking and running, hurdles, jumps, leaps, stepping stones, monkey bars, implicit quadruped gait transitions, moon gravity, push-responses, and more.

We present a technique for controlling physically simulated characters using user inputs from an off-the-shelf depth camera. Our controller takes a real-time stream of user poses as input, and simulates a stream of target poses of a biped based on it. The simulated biped mimics the user's actions while moving forward at a modest speed and maintaining balance. The controller is parameterized over a set of modulated reference motions that aims to cover the range of possible user actions. For real-time simulation, the best set of control parameters for the current input pose is chosen from the parameterized sets of pre-computed control parameters via a regression method. By applying the chosen parameters at each moment, the simulated biped can imitate a range of user actions while walking in various interactive scenarios.

Biped controller design pursues two fundamental goals; simulated walking should look human-like and robust against perturbation while maintaining its balance. Normal gait is a pattern of walking that humans normally adopt in undisturbed situations. It has previously been postulated that normal gait is more energy efficient than abnormal or impaired gaits. However, it is not clear whether normal gait is also superior to abnormal gait patterns with respect to other factors, such as stability. Understanding the correlation between gait and stability is an important aspect of biped controller design. We studied this issue in two sets of experiments with human participants and a simulated biped. The experiments evaluated the degree of resilience to external pushes for various gait patterns. We identified four gait factors that affect the balance-recovery capabilities of both human and simulated walking. We found that crouch gait is significantly more stable than normal gait against lateral push. Walking speed and the timing/magnitude of disturbance also affect gait stability. Our work would provide a potential way to compare the performance of biped controllers by normalizing their output gaits and improve their performance by adjusting these decisive factors.

We present a biped locomotion controller for humanoid models actuated by more than a hundred Hill-type muscles. The key component of the controller is our novel algorithm that can cope with step-based biped locomotion balancing and the coordination of many nonlinear Hill-type muscles simultaneously. Minimum effort muscle activations are calculated based on muscle contraction dynamics and online quadratic programming. Our controller can faithfully reproduce a variety of realistic biped gaits (e.g., normal walk, quick steps, and fast run) and adapt the gaits to varying conditions (e.g., muscle weakness, tightness, joint dislocation, and external pushes) and goals (e.g., pain reduction and efficiency maximization). We demonstrate the robustness and versatility of our controller with examples that can only be achieved using highly-detailed musculoskeletal models with many muscles.

We present a dynamic controller to physically simulate under-actuated three-dimensional full-body biped locomotion. Our data-driven controller takes motion capture reference data to reproduce realistic human locomotion through realtime physically based simulation. The key idea is modulating the reference trajectory continuously and seamlessly such that even a simple dynamic tracking controller can follow the reference trajectory while maintaining its balance. In our framework, biped control can be facilitated by a large array of existing data-driven animation techniques because our controller can take a stream of reference data generated on-the-fly at runtime. We demonstrate the effectiveness of our approach through examples that allow bipeds to turn, spin, and walk while steering its direction interactively.