A humanoid has infinitely many ways to stand up from sitting while maintaining balance, making sit-to-stand (STS) a challenging control problem. We synthesise natural humanoid STS motion from scratch using reinforcement learning, without demonstrations or reference trajectories. A single Proximal Policy Optimisation policy learns smooth, human-like rising driven by three complementary components. (i) A coupled force/chair-height curriculum is used. A vertical pelvis-assist force aids early trajectory exploration and decays over training. Taller chairs are unlocked with decaying assisting force. This ensures that the policy masters a viable STS trajectory at each chair height before being exposed to harder ones, avoiding the premature distribution shift that otherwise collapses generalisation. (ii) Motion robustness is achieved by randomly sampling from a large number of inverse kinematics-generated initial and target poses spanning over eight chair heights. (iii) A set of rewards is defined inspired from biomechanics and optimal control studies. They shape the robot's angular momentum for seat-off, and enable support-region transition via centre of pressure attraction function to ensure smooth low-effort actuation. On a deterministic force-free evaluator, the policy attains more than 97% balanced-standing success across eight chair heights. The policy generalises smooth motion across chair heights and enables the robot to rise from substantially deep-seated postures as compared to the state of the art.
Figures & tables
Assist force range
Unlocked chair lanes
Fmax>100 N
{0}
80<Fmax≤100 N
{0,1}
60<Fmax≤80 N
{0,1,2,3}
40<Fmax≤60 N
{0,1,2,3,4,5}
20<Fmax≤40 N
{0,1,2,3,4,5,6,7}
Fmax≤20 N
{1,2,3,4,5,6,7}
Table 1: Assist-force and chair-height curriculum. Each lane is a discrete seat height (lane 0 lowest). New lanes are progressively unlocked as Fmax decays.
Term
Expression
Weight
Common (applied every step)
roll
exp(−10ϕ2)
1
waist yaw
exp(−10qw2)
1
foot contact
exp(−50(dℓ+dr))
1
spin
exp(−0.5(ωx2+ωz2))
1
symmetry
exp(−5∥esym∥2)
1
Table 2: Rewards and their weights. Notation: hs=1.3236 m, c : COP, cˉhull : support-hull centroid, τf : foot tilt, nb : non-foot brace contacts, esym : mirror error, Δa=at−at−1 , Δ2a=at−2at−1+at−2 , other symbols as in the text). † rise-low ( ρ<0.1 ) only, with cref=21(cb+cˉfeet) there and the hull centroid otherwise.
Run
Removed component
Success (%)
Fall (%)
Act. jit.
DoF jit.
Sτ
Sq
Energy (Joules)
tstand (s)
Full
none (reference)
97.8
2.2
1.535
0.105
1.00
0.963
115.6
3.54
Rise-frac-off
rise-fraction staging
0.0
59.3
2.522
0.059
1.00
0.187
94.80
N/A
COP-off
measured-COP shaping
97.0
3.0
3.35
0.103
1.00
0.990
115.2
3.60
Schedule-off
progressive chair unlock
85.0
15.0
1.393
0.073
1.00
0.807
116.9
3.50
Omega-off
sagittal pitch-momentum
88.0
11.8
0.64
0.045
1.00
0.882
115.5
3.66
Table 3: Ablation study: each ablation omits one system component. All policies are evaluated deterministically (force-free, 8 lanes, 50 resets per lane).
Figure 1: Learned STS motion and randomized initial poses.
Figure 2: Temporal variation of ( 2(a) ) pitch joint torques and ( 2(b) ) angles during the rising phase. The red vertical line ( ≈0.5 s) indicates seat-off while the green dashed vertical line ( ≈3.9 s) marks 90% rise completion
Figure 3: The robot transfers its balance (quantified via its COP location) from the chair (encircled point) to the centroid of the foot-only support region.
This paper presents a reference-guided reinforcement learning framework to generate stand-up motion for a 29-DOF Unitree G1 humanoid on deformable soft ground, using a human demonstration recorded on hard ground. The terrain compliance is modelled using solref and solimp parameters from MuJoCo's rigid body soft-contact model. The rewards consists of (i) reference motion tracking through residual joint-position control and (ii) explicit recovery objectives such as pelvis height, torso uprightness, and the final posture. First, the policy is trained with the specified rewards considering hard ground. Next, the terrain stiffness is lowered by updating solref and the nominal surface penetration zone is expanded using solimp. Subsequent training enables the policy to adapt to the delayed support force generation due to significant surface penetration during contact-intensive phases while preserving the original demonstration pattern. The learned policy successfully completes the fallen-to-standing task in simulation, reaching the targeted pelvis height and uprightness, with a maximum contact penetration of approximately 40 mm during the process. The proposed method is demonstrated on two stand-up sequences and successfully achieves the final recovery objective on both hard and soft ground. Ablation studies show that reference tracking alone is insufficient for successful stand-up, and that explicit recovery rewards are essential.
Robotic assistance for human mobility could benefit care recipients and reduce caregiver burden during physically demanding tasks such as Sit-to-Stand (STS) and Lift-and-Transfer (L&T). Learning robot policies for these whole-body assistive tasks is challenging: real-world data collection for human-robot interaction is costly and safety-critical, making simulation an attractive alternative. Prior work, however, typically treats the human as static or minimally responsive and thus fails to capture the cooperative nature of mobility assistance. We present a simulation-based framework for learning whole-body assistive robot policies. At its core is a human policy trained with an adversarial motion prior that produces realistic care-recipient motion in response to the robot's actions. A human-adaptive scripted controller interacts with this policy to collect diverse robot demonstrations at scale. From these demonstrations, we train 3D diffusion policies that map segmented human point clouds and robot joint positions to whole-body actions for a mobile manipulator, without relying on privileged human state at deployment. The resulting robot policies successfully perform both STS and L&T assistance in simulation. Preliminary experiments further demonstrate sim-to-real transfer for STS after fine-tuning on a small amount of real-world data. Videos are available on our project page: https://rchi-lab.github.io/human-gen-mobility-assistance/
Shikun Ban, Shohei Nagai, Qingyang Song +4
The Robotics Institute, School of Computer Science, Carnegie Mellon University Pittsburgh, USA · Intuitive Autonomy San Jose, USA
Despite recent advances in humanoid locomotion, controllers optimized for command tracking and robustness tend to produce mechanical gaits, whereas controllers tied to human motion data often fail to generalize to commands outside the data distribution. This work introduces a learning framework that balances these competing objectives to synthesize real-time steerable, robust, and biomimetic locomotion policies from human data. Using an in-house curated locomotion dataset covering diverse speeds and directions, we first learn a natural locomotion prior policy through a teacher-student distillation process. Specifically, we train a full-body reference-conditioned policy with Reinforcement Learning (RL), then distill it into a lightweight prior policy conditioned solely on proprioception and a planar torso-velocity steering command. Next, we fine-tune the prior policy with multi-task RL to expand command coverage and robustness beyond the data distribution, pairing a goal-conditioned task that tracks arbitrary commands with a reference-guided task that tracks the human data as an explicit style regularizer. We validate our framework on three humanoid robots: the Boston Dynamics Atlas R1, Atlas D1, and Unitree G1. Experimental results demonstrate robust performance across real-world scenarios, including direct user-controlled locomotion in indoor and outdoor environments, and integration as the locomotion layer within hierarchical control stacks. Benchmarks against Tabula Rasa RL policies trained without human data and ablation studies confirm that our framework yields a lightweight, deployable policy that reconstructs coordinated whole-body behavior from a steering command, retaining the human gait characteristics while remaining robust and fully steerable.
Mike Zhang, Dongho Kang, Kevin Bergamin +19
RAI Institute, USA · Boston Dynamics, USA · Carnegie Mellon University, USA