SPOT: Spatio-Temporal Obstacle-free Trajectory Planning for UAVs in an Unknown Dynamic Environment

TL;DR

SPOT targets reactive UAV navigation in unknown, dynamic environments by operating in a four-dimensional state space $(x,y,z,t)$. It couples a spatio-temporal ST-RRT* planner with vision-based Safe Flight Corridor generation and MINCO-based trajectory optimization, augmented by a backup trajectory module to prevent deadlocks. The approach explicitly handles dynamic obstacles through time-augmented planning, obstacle prediction, and a runtime FSM that can switch to a safe backup path when needed. Extensive simulations and hardware experiments demonstrate improved robustness and safety over state-of-the-art methods, highlighting SPOT's mapless, perception-driven capabilities for real-world UAV operation.

Abstract

We address the problem of reactive motion planning for quadrotors operating in unknown environments with dynamic obstacles. Our approach leverages a 4-dimensional spatio-temporal planner, integrated with vision-based Safe Flight Corridor (SFC) generation and trajectory optimization. Unlike prior methods that rely on map fusion, our framework is mapless, enabling collision avoidance directly from perception while reducing computational overhead. Dynamic obstacles are detected and tracked using a vision-based object segmentation and tracking pipeline, allowing robust classification of static versus dynamic elements in the scene. To further enhance robustness, we introduce a backup planning module that reactively avoids dynamic obstacles when no direct path to the goal is available, mitigating the risk of collisions during deadlock situations. We validate our method extensively in both simulation and real-world hardware experiments, and benchmark it against state-of-the-art approaches, showing significant advantages for reactive UAV navigation in dynamic, unknown environments.

Figures (9)

• Figure 1: UAV navigation in a dynamic environment with backup planning. (a) The nominal trajectory (green) becomes unsafe due to dynamic obstacles. (b) A backup trajectory (red) is generated reactively to move the UAV toward a safer region. (c) The UAV tracks the backup trajectory, replanning it as needed to maintain safety. (d) Once a feasible path to the goal becomes available, the UAV switches back to goal-directed planning. Convex collision-free corridors are omitted for clarity. Obstacles within the UAV’s sensing range are shown with red bounding boxes, while those outside the range are shown with green bounding boxes.
• Figure 2: System architecture of SPOT for reactive UAV motion planning in dynamic, unknown environments.
• Figure 3: FSM governing the spatio-temporal RRT$^\star$ planning pipeline, operating in three modes: Initial, Incremental, and Backup.
• Figure 4: Spatio-temporal RRT path generated in a dynamic environment. The environment provides a point cloud of obstacles, propagated into the future using velocity information. Each sphere represents the safety margin of a node (feasibility set) at its time of arrival. A color gradient from red to blue illustrates the flow of time, from the present (red) toward the future (blue).
• Figure 5: Spatio-temporal convex collision-free polyhedra constructed along the path segment between nodes $n_i$ and $n_j$. The evolution of the safe corridor is color-coded from red to gray over the time span $[t_i, t_f]$. During the same interval, the predicted positions of dynamic obstacles are represented as point clouds, also color-coded from $t_i$ to $t_f$. These aggregated point clouds constitute the dynamic obstacle set $\mathcal{O}^d_{ij}$ used in corridor construction. Final optimized trajectory is visible in green.