A Tree-based Next-best-trajectory Method for 3D UAV Exploration

Abstract: This work presents a fully integrated tree-based combined exploration-planning algorithm: Exploration-RRT (ERRT). The algorithm is focused on providing real-time solutions for local exploration in a fully unknown and unstructured environment while directly incorporating exploratory behavior, robot-safe path planning, and robot actuation into the central problem. ERRT provides a complete sampling and tree-based solution for evaluating "where to go next" by considering a trade-off between maximizing information gain, and minimizing the distances travelled and the robot actuation along the path. The complete scheme is evaluated in extensive simulations, comparisons, as well as real-world field experiments in constrained and narrow subterranean and GPS-denied environments. The framework is fully ROS-integrated, straight-forward to use, and we open-source it at https://github.com/LTU-RAI/ExplorationRRT.

• The paper presents a novel tree-based ERRT algorithm that integrates information gain assessment with RRT* for optimized trajectory planning.
• The paper demonstrates that ERRT markedly improves 3D and subterranean exploration efficiency compared to conventional methods.
• The paper shows that incorporating NMPC for actuation trajectory computation enables UAVs to navigate complex, GPS-denied spaces with dynamic feasibility.

Overview of "A Tree-based Next-best-trajectory Method for 3D UAV Exploration"

The paper "A Tree-based Next-best-trajectory Method for 3D UAV Exploration" advances the domain of robotic exploration with the development of Exploration-RRT (ERRT), an integrated exploration-planning algorithm for unmanned aerial vehicles (UAVs). This approach is designed to function autonomously in unknown and unstructured environments, optimizing exploration by addressing the quintessential question of "where to go next." The work emphasizes seamless functionality in subterranean and GPS-denied spaces, showcasing the method's efficiency through substantial simulation and field experiments.

Central to the ERRT algorithm is the simultaneous consideration of information gain, travel distance, and robot actuation within an RRT-based framework. Traditional exploration methods often segregate navigation from path planning, whereas ERRT integrates these components, allowing it to deliver dynamically viable path suggestions that consider actuator limitations and maximize environmental comprehension via sensor data.

Key elements in the ERRT implementation include:

1. Candidate Goal Generation: Employing Poisson disk sampling, ERRT identifies candidate exploration targets ensuring information gain greater than zero.
2. RRT Structure and Path Optimization: Utilizing RRT$^*$ for tree expansion, the method enhances trajectories through iterative collision checks, determining the shortest viable paths.
3. Actuation Trajectory Computation: Leveraging nonlinear model predictive control (NMPC), ERRT computes paths with feasible dynamic properties for UAVs, emphasizing smooth transitions.
4. Information Gain Assessment: The algorithm evaluates information gain continuously along the trajectory rather than solely at endpoints, offering a nuanced understanding of potential exploration benefits.

The framework demonstrates superior adaptability and performance when benchmarked against established methods such as Graph-Based Planner (GBP) and the Rapid Exploration Framework (REF) in simulation environments. ERRT's capability of efficiently exploring large 3D and complex subterranean environments highlights its advantage in scenarios demanding rapid exploration decisions within constrained timeframes.

Implications and Future Directions

The ERRT proposition has significant implications for autonomous exploration in challenging environments such as subterranean mining operations, search-and-rescue missions, and planetary exploration. By effectively managing computational complexity and dynamically integrating exploration and planning, ERRT enhances the real-time decision-making capabilities of UAVs, thus broadening their operational scope.

Future research could focus on further reducing the computational demands through smarter sampling techniques, possibly integrating machine learning models for non-uniform sampling. Another promising avenue would be to incorporate localization constraints directly into the trajectory planning process, thus improving UAV performance in GPS-denied environments where SLAM challenges predominate.

Moreover, expanding ERRT's capabilities to address complete coverage strategies and implementing a global exploration framework will bolster its utility in persistent, long-duration missions where maintaining operational efficacy across exploratory phases is critical.

In conclusion, ERRT presents a robust method for solving the exploration-planning problem by seamlessly integrating multiple complex factors, delivering a system that not only advances the state of autonomous UAV exploration but also sets a solid foundation for subsequent enhancements in autonomous navigation technologies.