ResearchBenchmarksReaching on Kinematic-simulation n=1000 targetsFollow100Success Rate (%)Min-jerk68.69676.82384.9593.077Jun 14, 2026Evaluation ResultsMethodMethodLinksSuccess Rate (%)Energy Consumption (J) (All Episodes)Energy Consumption (J) (Successes Only)Episode Duration (s)Final Position Error (cm)Final Orientation Error (°)Smoothness (MSJ) (rad² s⁻⁶)Action Saturation (%)Min-jerkpolicy=closed-form tra...policy=closed-form trajectory, dp tolerance=4 cm, do tolerance=8.6°2026.061002221.951.276—7.88212.1RL (SAC)policy=deterministic,...policy=deterministic, checkpoint=5 × 10⁶ steps, dp tolerance=4 cm, do tolerance=8.6°2026.0669.9163.798.161.542—16.491.3124.6