Humanoid-Gym: What a Second Simulator Can Check Before Hardware
Export a Humanoid-Gym policy from Isaac Gym and replay it in MuJoCo to locate interface and physics mismatches. A second-engine pass is a screen, not hardware proof.
Physical AI Lab
OpenVLA-OFT: What Parallel Action Chunks Change at Deployment
OpenVLA-OFT generates continuous action chunks in parallel. Check chunk timing, action units, inputs, compute and multimodal-task needs on the target robot before choosing it.
ASAP: Using Real Trajectories to Correct Simulation Training
ASAP fits a residual from real humanoid rollouts, uses it to fine-tune a policy in simulation, then deploys the policy alone. Trace each evidence boundary before transfer.
RMA: Adapting a Walking Policy from Recent Robot History
RMA estimates a latent adaptation signal from recent states and actions while keeping deployed policy weights fixed. Evaluate the transition after a terrain or payload change.
DINO-WM: Planning Actions in Visual Feature Space
DINO-WM predicts action-conditioned visual features and plans toward a goal image without reconstructing every candidate frame. Compare rollout accuracy and executed success together.
NoMaD: Switching Between Exploration and Visual Goal Reaching
NoMaD’s goal mask switches a shared visual policy between undirected exploration and goal-image pursuit. The mission layer still selects subgoals and decides when to switch.
ViNT: Adapting Visual Navigation to a Different Goal Interface
ViNT maps image goals, local coordinates and route commands into a navigation policy’s goal interface. Compare labels and errors by modality before using a new route source.
GNM: Making Navigation Trajectories Comparable Across Robots
GNM-style cross-robot navigation needs aligned camera history, action scale and robot context. Audit each trajectory before mixing data, then test transfer by platform.
Nav2 RateController: Why Running Actions Still Receive Ticks
Nav2 RateController keeps ticking a running child between nominal replanning intervals. Distinguish tree ticks, action requests and completed paths before tuning hz.
Nav2 Velocity Smoother: Choosing Command or Odometry Feedback
Choose Nav2 velocity-smoother feedback by comparing the last commanded speed with fresh odometry at the smoothing cycle. Check tracking, delay and stale-command behavior.