DART algorithm lands 30/30 off-road jumps, beating 0/30 baseline

A team led by Yu Hu from (likely) a Chinese university has published a robotics paper on arXiv introducing DART (Dual-Axis Airborne Reachability-Gated Torque-Reaction), a control system for off-road vehicles that become airborne after cresting steep lips, ledges, or ditches. The central insight is that mid-air control is brutally limited: on a 1383kg vehicle, the wear-induced angular momentum of the wheels can only recover 9-13°/s of nose-up pitch-rate change under typical takeoff speeds (double that if you aggressively brake the wheels). Pushing the drivetrain to its hard limits raises the ceiling to just 16-18°/s. Anything beyond that is physically unrecoverable once the vehicle leaves the ground, so any meaningful safety must come before takeoff.

DART therefore back-propagates the landing constraint into a closed-form 'feasible-takeoff set,' providing a conservative go/no-go condition and a pre-takeoff speed-shaping law. In flight, it regulates pitch and roll using steer-resolved wheel-reaction torque, gated by a roll latch derived from yaw-coupling analysis. In deterministic full-scale simulation, the pre-takeoff regulator reduces touchdown speed by 36% and lifts perfect on-target landings from 0/30 to 30/30. On steep-lip approaches, DART's airborne law completes 29/30 safe landings versus 0/30 for reaction-wheel-style PD and time-optimal bang-bang controllers. On banked run-ups it holds median pitch error under 2° across all cross-slopes. All results are simulation-based; hardware validation remains open.