DYNUS: Uncertainty-aware Trajectory Planner in Dynamic Unknown Environments (2504.16734v3)

Abstract: This paper introduces DYNUS, an uncertainty-aware trajectory planner designed for dynamic unknown environments. Operating in such settings presents many challenges -- most notably, because the agent cannot predict the ground-truth future paths of obstacles, a previously planned trajectory can become unsafe at any moment, requiring rapid replanning to avoid collisions. Recently developed planners have used soft-constraint approaches to achieve the necessary fast computation times; however, these methods do not guarantee collision-free paths even with static obstacles. In contrast, hard-constraint methods ensure collision-free safety, but typically have longer computation times. To address these issues, we propose three key contributions. First, the DYNUS Global Planner (DGP) and Temporal Safe Corridor Generation operate in spatio-temporal space and handle both static and dynamic obstacles in the 3D environment. Second, the Safe Planning Framework leverages a combination of exploratory, safe, and contingency trajectories to flexibly re-route when potential future collisions with dynamic obstacles are detected. Finally, the Fast Hard-Constraint Local Trajectory Formulation uses a variable elimination approach to reduce the problem size and enable faster computation by pre-computing dependencies between free and dependent variables while still ensuring collision-free trajectories. We evaluated DYNUS in a variety of simulations, including dense forests, confined office spaces, cave systems, and dynamic environments. Our experiments show that DYNUS achieves a success rate of 100% and travel times that are approximately 25.0% faster than state-of-the-art methods. We also evaluated DYNUS on multiple platforms -- a quadrotor, a wheeled robot, and a quadruped -- in both simulation and hardware experiments.

Overview of DYNUS: Uncertainty-aware Trajectory Planner in Dynamic Unknown Environments

The paper introduces DYNUS, a sophisticated trajectory planning system that handles navigation through dynamic and unknown environments. The planning of trajectories in such settings is a complex challenge, primarily because the paths of obstacles cannot be predicted accurately, necessitating continuous replanning to avoid collisions. The DYNUS framework addresses these challenges by combining several key components that ensure efficient and safe navigation.

Key Contributions

1. Global Planner (DGP): DYNUS utilizes a hybrid global planner that combines the strengths of Jump Point Search (JPS) and Dynamic A*. JPS rapidly computes a path by considering static obstacles, while Dynamic A* accounts for dynamic ones, optimizing the path calculation process in terms of both time and adaptability.
2. Safe Local Planning Framework: This framework utilizes exploratory, safe, and contingency trajectories. Exploratory trajectories are planned forward into unknown areas, whereas safe trajectories provide alternative routes. Contingency trajectories ensure immediate safety in case of unexpected dynamic obstacle behavior.
3. Variable Elimination for Faster Computation: A novel technique in the optimization process significantly reduces the number of decision variables, allowing faster computation. This is critical in maintaining real-time responsiveness without compromising safety.
4. Efficient Yaw Optimization: The yaw optimization process actively balances the need to track dynamic obstacles with maintaining a view in the direction of motion. This is achieved through a graph search paired with B-spline fitting, ensuring smooth transitions and effective obstacle monitoring.
5. Robust Dynamic Obstacle Tracking: Utilizing an Adaptive Extended Kalman Filter (AEKF), DYNUS dynamically adjusts process and measurement noise covariances to maintain accurate tracking of dynamic obstacles.

Numerical Results

DYNUS demonstrated a 100% success rate in numerous challenging simulations, including dense forests, office environments, and cave systems. It consistently operated faster than comparable systems with a reduction in travel time by approximately 25.0% compared to state-of-the-art methods. This success is largely attributed to its ability to rapidly compute safe trajectories and efficiently replan in real-time.

Practical and Theoretical Implications

The practical implications of DYNUS are significant in realms where autonomous navigation in dynamic environments is essential, such as urban UAV applications, exploration in confined spaces, and rescue missions. Theoretically, DYNUS advances current understanding of trajectory planning by incorporating adaptive mechanisms for real-time obstacle tracking and prediction, which could stimulate further research into robust optimization techniques and dynamic environment modeling.

Future Research Directions

Future developments could focus on large-scale deployment of DYNUS-like systems across varied terrains and conditions. Enhancing computational efficiency, such as exploring more advanced variable elimination methods or integrating machine learning models for obstacle behavior prediction, could further improve performance. Additionally, expanding the system to handle adversarial dynamic obstacles presents an interesting research challenge.

In conclusion, DYNUS represents a significant step forward in trajectory planning for dynamic and unknown environments, embodying a strategic blend of innovative algorithms that pave the way for more autonomous navigation solutions. Its success in both simulations and hardware experiments highlights the robustness and adaptability of the system, marking it as a valuable contribution to autonomous systems research.