Humanoid-Actuator-ThermalEnergetics

Physics-informed (Joule heating + first-order thermal RC model) humanoid walking-gait dataset: 18 robot embodiments, 576 trials, per-joint torque/current/winding-temperature trajectories, with thermal derating and shutdown outcome labels -- a real deployment constraint most gait datasets ignore.

About Dataset

Humanoid-Actuator-ThermalEnergetics simulates walking-gait trials with full per-joint actuator electrical and thermal state, across 18 independently-sampled robot embodiments (torque constant, winding resistance, thermal resistance/capacitance, thermal warning/cutoff temperatures, peak torque). Each of 5 leg joints (hip pitch/roll, knee pitch, ankle pitch/roll) follows a gait-cycle torque profile scaled by walking speed, terrain incline, and carried payload. Torque converts to motor current via the standard torque-constant relationship, and resistive (I-squared-R) heating drives a first-order thermal model of winding temperature -- the same lumped-parameter model used in real motor datasheets and thermal-protection firmware. Once winding temperature exceeds a per-actuator warning threshold, commanded torque is progressively derated to slow further heating, mirroring real thermal-protection behavior. Trials are labeled with one of three outcomes: completed, thermal_derate_triggered, or thermal_shutdown. Each timestep records 22 channels: per-joint torque/current/winding-temperature (15 channels across 5 joints) plus walking speed, terrain incline, payload, ambient temperature, gait-cycle phase, and a derating-active flag. This is a fully synthetic dataset -- no real robot telemetry, proprietary, or personal data was used.

Purpose of Dataset

This Dataset Supports Research On Predictive Maintenance And Thermal-aware Control For Humanoid Actuators -- Specifically, Predicting Motor Overheating Before It Forces A Shutdown, Which Is A Real Operational Failure Mode That Ends A Robot's Task Even When Its Balance And Control Are Otherwise Fine. It Targets Researchers And Engineers Building Thermal-derating-aware Gait Planners, Actuator Health-monitoring Systems, Or Early-warning Overheating Predictors, Particularly Across A Fleet Of Robots With Different Actuator Designs Rather Than One Fixed Hardware Platform. The Embodiment-grouped Design Supports Studying Whether A Thermal-shutdown Predictor Trained On Some Actuator Designs Transfers To New Hardware It Has Never Run On.