Autonomous vehicles track planned trajectories close to obstacles. Unknown disturbances make them deviate from these trajectories. With a bound on the tracking error that disturbances can cause, a planned trajectory can be certified as safe before execution. Existing set-based methods bound the tracking error by one invariant ellipsoid for the whole trajectory or by intervals, which are both conservative. For aerial and space vehicles, the tracking error evolves on a matrix Lie group with nonlinear dynamics. However, for group-affine systems, the error dynamics are linear in Lie algebra coordinates when the vehicle receives the inputs of the reference. The control input and the disturbance enter through the inverse right Jacobian, a matrix that depends on the error. We propagate the linear part of the error dynamics with zonotopes and enclose the effect of the inverse right Jacobian on the disturbance in closed form. The resulting tubes are tight and provably contain all reachable tracking errors in continuous time. When some error components have no input, the inverse right Jacobian leaves unmatched nonlinear terms in the closed loop. None of the tested methods certifies such a closed loop. Hybrid coordinates make these terms matched, so that our controller cancels every nonlinear term that does not contain the disturbance. Our tubes give tighter bounds than the ellipsoid method in simulations of a planar aircraft and a fully actuated multirotor and in real flights of a quadrotor, where the tube contains every measured attitude error. Under larger disturbances and the same feedback gain, the ellipsoid method yields no valid ellipsoid, whereas our tubes remain valid.
Figures & tables
Figure 1. The tracking error on the Lie group and in Lie algebra coordinates. In the left panel, the vehicle X(t) follows the reference Xˉ(t) , and the tube contains every tracking error that the disturbances can cause. A tube that avoids the obstacle certifies the reference. The inset shows a zonotope mapped onto SO(3) by the exponential map, whose edges become curved, so we propagate the zonotopes in Lie algebra coordinates. The right panel shows zonotopes in Lie algebra or hybrid coordinates, which contain the tracking error at the time steps, and the tube, whose time-interval sets contain it between the time steps. The dashed curves are a trajectory of the vehicle and its tracking error. The enlarged zonotope consists of its center c and its generators, and each step adds generators for the disturbance. The bounds on the attitude error and the position error are read from these sets.
Figure 2. One step of Algorithm 1 , with illustrative sets. The lower row is the solution ( 18 ) of the closed-loop error equation over the step. The linear part maps the zonotope to another zonotope without over-approximation, the input adds a zonotope with one generator per disturbance component, and a box encloses the remaining terms. The sum of the three sets contains the tracking error at the end of the step.
Figure 3. Tubes of the planar aircraft in the world frame around an obstacle near the reference path. (a) The whole path. (b) Under the original disturbance, the tube clears the obstacle, whereas the region swept by the ellipsoid overlaps it. (c) Under the strong disturbance, the tube overlaps the obstacle, and so do trajectories of the true closed loop. No ellipsoid is valid.
method
w1=1 heading [rad]
w1=1 position [m]
time
w1=5 heading [rad]
w1=5 position [m]
time
lower bound
0.112
1.15
—
0.496
5.71
—
tube, refined steps
0.131
1.28
12 s
0.757
8.04
29 s
tube, fixed step
0.134
1.30
1.2 s
0.849
8.96
1.2 s
per-channel ellipsoid
0.268
1.59
4.2 s
no valid set
2.0 s
joint-bound ellipsoid
2.28
10.7
32 s
no valid set
0.9 s
Table 1. Maximum over [0,T] of the bounds on the heading error and on the position error of the planar aircraft, with the computation time. The ratios to the lower bound in the text are taken at the final time T .
method
nominal, attitude [rad]
nominal, position [m]
time
strong, attitude [rad]
strong, position [m]
time
lower bound
0.134
0.701
—
0.419
2.22
—
tube
0.167
0.898
12 s
0.630
3.21
34 s
ellipsoid
1.05
1.99
1.9 min
no valid set
44 s
time-varying norm bound
0.964
2.00
3.0 min
3.02
6.27
3.0 min
Table 2. Maximum over [0,T] of the bounds on the attitude error angle and on the position error of the multirotor, with the computation time. The ellipsoid is the joint-bound ellipsoid, which gives the smaller bounds. In every step, the tube keeps the box input or the ball input of Lemma 6 , whichever gives the smaller attitude bound.
Figure 4. Bounds on the attitude error and the position error of the multirotor under nominal and strong disturbances. The tube lies just above the largest error of the true closed loop and far below the ellipsoid and the time-varying norm bound. Under the strong disturbances no ellipsoid is valid.
Figure 5. (a) The quadrotor. (b), (c) Magnitudes of the roll and yaw components of the attitude error measured in eight flights, with the bounds of the tube and the ellipsoid. The tube contains every measurement and is tighter than the ellipsoid.
Discrete trajectory checks can miss collisions between sampled robot states. Reachability-based certification bounds motion between states, but uniform time partitions waste computation where clearance is large. We present lazy adaptive reachability certification (LARC), which checks a planned trajectory by bisecting only intervals with an inconclusive clearance test. For piecewise-cubic Hermite joint trajectories, the method bounds link occupancy using midpoint capsules inflated by exact componentwise speed maxima. Certified intervals covering the trajectory provide continuous-time external-obstacle clearance, subject to geometric containment, static obstacles, and a prescribed margin. On 160 AgileX PIPER trajectories from 80 start-goal pairs, LARC matched all decisions of the fixed-fine baseline at depth nine. It used 20328 interval evaluations (24.8% of baseline work), with a median paired speedup of 10.28x. A separate MoveIt/FCL audit checked 158051 states and detected collisions in 21 direct-interpolation controls, none of which LARC certified. The method reduced computation under a shared certificate model, but 27 of 139 sampled-clear trajectories remained uncertified. The sampled audit cannot independently prove continuous-time clearance.
Yu Feng, Hao Wu, Yuzhe Wang +1
Department of Mechanical Engineering, National University of Singapore, Singapore · Singapore Institute of Manufacturing Technology, Agency for Science Technology and Research, Singapore
This paper presents a systematic framework for computing formally guaranteed trajectory tracking error bounds for autonomous helicopters based on Robust Positive Invariant (RPI) sets. The approach establishes a closed-loop translational error dynamics which is cast into polytopic linear parameter-varying form with bounded additive and state-dependent disturbances. Ellipsoidal RPI sets are computed, yielding explicit position error bounds suitable as certified buffer zones in upper-level trajectory planning. Three controller architectures are compared with respect to the conservatism of their error bounds and tracking performance. Simulation results on a nonlinear helicopter model demonstrate that all architectures satisfy the derived bounds, while highlighting trade-offs between performance and the conservatism of the computed invariant set.
Philipp Schitz, Johann C. Dauer, Paolo Mercorelli
Institute of Flight Systems, German Aerospace Center (DLR), 38108 Braunschweig, Germany · Institute for Production Technology and Systems, Leuphana University of Lueneburg, 21335 Lueneburg, Germany
Reachability-based Trajectory Design (RTD) is a provably safe, real-time trajectory planning framework that combines offline reachable-set computation with online trajectory optimization. However, standard RTD implementations suffer from two key limitations: conservatism induced by worst-case reachable-set overapproximations, and an inability to account for real-time disturbances during execution. This paper presents RTD-RAX, a runtime-assurance extension of RTD that utilizes a non-conservative RTD formulation to rapidly generate goal-directed candidate trajectories, and utilizes mixed monotone reachability for fast, disturbance-aware online safety certification. When proposed trajectories fail safety certification under real-time uncertainty, a repair procedure finds nearby safe trajectories that preserve progress toward the goal while guaranteeing safety under real-time disturbances.
Evanns Morales-Cuadrado, Long Kiu Chung, Shreyas Kousik +1
S. Coogan is also with the School of Civil and Environmental Engineering.