Autonomous UAV flight through cluttered and partially unknown environments requires reasoning not only about observed obstacles but also about occluded regions that the sensor cannot observe. We present OA-MPPI, an obstacle- and occlusion-aware extension of Model Predictive Path Integral (MPPI) control for quadrotor flight that accounts for potential moving agents emerging from these regions into the vehicle's path. At every planning step, we extract a 3D occlusion boundary from the online occupancy map and use it to model the regions that hidden agents could reach over the prediction horizon. We penalize trajectories that enter these expanding regions within MPPI rollouts generated using nonlinear quadrotor dynamics and accounting for individual rotor thrust limits. We validate the proposed approach in simulation and hardware flight experiments, with the complete pipeline running onboard the vehicle in real time. Results show increased clearance from occlusion boundaries compared to baseline MPPI in both settings, as well as avoidance of an agent emerging from occlusion in simulation.
Figures & tables
Fig. 1: Example of OA-MPPI in operation. Top: experimental setup. Bottom: occupancy map and sampled rollouts. Red voxels denote occupied space, and orange voxels mark the extracted occlusion boundary. Green lines indicate collision-free rollouts, yellow lines indicate rollouts entering an expanding keep-out region, and red lines indicate rollouts colliding with mapped obstacles. The green sphere marks the goal. Ground voxels are omitted in the plot for clarity.
Fig. 2: Overview of the OA-MPPI framework. “Ref.” denotes the reference trajectory sent to the controller.
MPPI param.
Cost definition constants
H
30
wgoal
0.1
K
500
wgoalterm
5.0
λ
0.1
wcol
50.0
Δt
0.1s
wvel
15.0
Σ
diag(0.60,0.15,0.15,0.05)
R
diag(0.01,0.05,0.05,0.10)
TABLE I: MPPI horizon/sampling parameters and cost weights used.
Fig. 3: Occlusion boundary extraction for three obstacles surrounding the drone. Top: top-down view; blue line point to the nearest occlusion voxel (red), with sphere color indicating the keep-out radius at increasing horizon times, from t=0 s (blue) to t=3 s (full horizon H , red). Bottom: 3D view; yellow lines mark the occlusion lines of sight, with the expanding keep-out region (green concentric spheres) around the closest boundary point.
Fig. 4: Simulation of an agent emerging from occlusion. The drone collides with the agent under baseline MPPI (gray), while OA-MPPI (blue) avoids the collision. The agent is shown as a purple sphere and speed 0.4 m/s. Top: t=2.8 s; bottom: t=3.9 s.
Fig. 5: Top: simulation setup for flight over an occluding box. Bottom: side view in the xz -plane at t=2.1 s. OA-MPPI (blue) takes a higher path over the box than baseline MPPI (gray), increasing clearance from the occlusion boundary.
Fig. 6: Experiments across the three test scenarios: single wall (left), pillar (middle), and two-wall corridor (right). Top: physical setup, with the drone flying from its start position toward the goal near or behind the occluding structure(s). Bottom: top-down occupancy map for the same run, showing the occlusion-agnostic (gray) and occlusion-aware (blue) trajectories from start (green dot) to goal (yellow star), with concentric contours marking the time-growing keep-out radius rkeep(t) around the occlusion boundary and dashed orange lines indicating the occlusion lines of sight. The drone shape is only indicated on the occlusion-agnostic line for clarity. The goal is reached if within 0.5 m from it.
Scenario
Method
Time [ s ]
Dist. [ m ]
Vel. [ ms−1 ]
Single wall
MPPI
2.49
2.93
1.18
OA-MPPI
4.19
3.76
0.91
Pillar
MPPI
3.58
3.77
1.11
OA-MPPI
3.74
4.21
1.14
Two walls
MPPI
3.06
3.52
1.15
OA-MPPI
3.75
3.85
1.03
TABLE II: Comparison of baseline MPPI and OA-MPPI across three flight scenarios.
This work proposes the L1 Adaptive Model Predictive Path Integral (L1-MPPI). It cascades L1 adaptive control with the Model Predictive Path Integral (MPPI) to improve tracking of high-speed UAV trajectories. Thanks to the L1augmentation, the tracking remains accurate even under model uncertainties and external disturbances, such as an additional payload or a mismatch in the modeled aerodynamic drag. In contrast to existing MPPI approaches for UAV control that do not explicitly model aerodynamic effects, varying payloads, and typically neglect the dynamics of low-level motor controllers, our L1-MPPI approach enhances the dynamic model used in the MPPI by incorporating the low-level flight controller and motor dynamics, as well as an iterative mixing scheme that reflects the approach of the low-level controller. The proposed method demonstrates improved tracking performance in both simulation and the real world, even when the UAV is subjected to an unknown payload. In flight with 35% mass increase, our approach lowers the RMSE by 58.61% with respect to plain MPPI. Compared to the same MPPI using an online mass estimator in place of the L1 augmentation, the RMSE is lower by 38.59%. During the real-world experiments the UAV reaches speeds up to 13.50 m/s and accelerations up to 2.5 g.
Lukáš Kotek, Ondřej Procházka, Vojtěch Vonásek +2
Multi-robot Systems Group, Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic
This paper presents a new robust integrated planning and control (IPC) strategy for multirotor uncrewed aerial vehicles. We propose a nonlinear model predictive control (NMPC) formulation that embeds control barrier functions (CBFs) as exponential penalties, improving feasibility while ensuring smooth obstacle avoidance under tight input bounds. The penalty weights provide a practical tuning knob to trade off tracking accuracy against avoidance aggressiveness. We enhance the system robustness by employing a high-gain disturbance observer (HGDO) to estimate and compensate for external disturbances. We also incorporate a Kalman filter (KF) for computationally efficient, real-time prediction of obstacle motion, enabling avoidance of moving obstacles. Comparative studies against both conventional NMPC and NMPC with hard CBF constraints, validated in Gazebo and hardware experiments, demonstrate superior feasibility, safety, and robustness. To the best of our knowledge, this is the first hardware-validated NMPC-CBF IPC framework, offering a practical step toward safe quadrotor deployment in dynamic environments.
Zeinab Shayan, Mohammadreza Izadi, Reza Faieghi
Autonomous Vehicles Laboratory, Department of Aerospace Engineering, Toronto Metropolitan University, Toronto, Canada.
With the widespread availability of parallel computing hardware, sampling-based motion planning methods such as Model Predictive Path Integral (MPPI) control have become increasingly powerful for complex nonlinear systems in non-smooth task spaces. However, the sampling and forward-simulation pipeline in MPPI suffers from averaging-induced failure in cluttered environments, where the importance-weighted update averages incompatible rollouts and leads to hesitation or even collision when an obstacle lies directly ahead. This paper proposes Clustering-Embedded MPPI (CE-MPPI), a framework that architecturally resolves the averaging-induced failures inherent in standard MPPI within non-convex environments. Rather than simply mitigating interference, CE-MPPI redefines the control law by integrating a high-fidelity pruning and clustering stage. By leveraging density-based spatial clustering of applications with noise (DBSCAN) alongside a novel geometric direction feature that is extracted from collision-derived reference points, the system isolates feasible trajectory modes from the noise of infeasible rollouts. This is paired with an intelligent selection logic that optimizes for minimum cost in static scenes while actively steering opposite to obstacle flux in dynamic environments. Experiments in 2-D JAX-accelerated simulations show that CE-MPPI alleviates obstacle-front hesitation and avoids persistent coupling with moving obstacles in dynamic scenes. In particular, real-world tests on a 6-DoF UR5e manipulator with CUDA-parallel rollouts in Isaac Gym achieve a 48% reduction in time-to-goal and a 12% shorter end-effector path.
Zidong Liu, Kaixin Chang, Xu Chen
Department of Mechanical Engineering, University of Washington, Seattle, WA, USA.