New method certifies safe paths for autonomous vehicles
A new way to bound tracking errors helps aerial and space vehicles stay near planned trajectories.
Autonomous vehicles track planned trajectories near obstacles. Unknown disturbances can cause these vehicles to deviate from their paths.
A planned trajectory can be certified as safe before execution if there is a bound on the tracking error. Existing methods use one invariant ellipsoid or intervals to bound error, but these are conservative.
For aerial and space vehicles, error evolves on a matrix Lie group with nonlinear dynamics. The error dynamics are linear in Lie algebra coordinates for group-affine systems when the vehicle receives the reference inputs.
The method propagates the linear part of the error dynamics with zonotopes. It encloses the effect of the inverse right Jacobian on the disturbance in closed form. The resulting tubes are tight and contain all reachable tracking errors in continuous time.
Hybrid coordinates make unmatched nonlinear terms matched. This allows the controller to cancel every nonlinear term that does not contain the disturbance.
Simulations of a planar aircraft and a fully actuated multirotor showed tighter bounds than the ellipsoid method. In real flights of a quadrotor, the tube contained every measured attitude error. Under larger disturbances, the ellipsoid method yields no valid ellipsoid, but the tubes remain valid.
Key facts
- The resulting tubes are tight and provably contain all reachable tracking errors in continuous time.
- The method provides tighter bounds than the ellipsoid method in simulations and real quadrotor flights.
- The controller cancels every nonlinear term that does not contain the disturbance using hybrid coordinates.