Teach-Repeat-Replan: A Complete and Robust System for Aggressive Flight in Complex Environments (1907.00520v1)

Abstract: In this paper, we propose a complete and robust motion planning system for the aggressive flight of autonomous quadrotors. The proposed method is built upon on a classical teach-and-repeat framework, which is widely adopted in infrastructure inspection, aerial transportation, and search-and-rescue. For these applications, human's intention is essential to decide the topological structure of the flight trajectory of the drone. However, poor teaching trajectories and changing environments prevent a simple teach-and-repeat system from being applied flexibly and robustly. In this paper, instead of commanding the drone to precisely follow a teaching trajectory, we propose a method to automatically convert a human-piloted trajectory, which can be arbitrarily jerky, to a topologically equivalent one. The generated trajectory is guaranteed to be smooth, safe, and kinodynamically feasible, with a human preferable aggressiveness. Also, to avoid unmapped or dynamic obstacles during flights, a sliding-windowed local perception and re-planning method are introduced to our system, to generate safe local trajectories onboard. We name our system as teach-repeat-replan. It can capture users' intention of a flight mission, convert an arbitrarily jerky teaching path to a smooth repeating trajectory, and generate safe local re-plans to avoid unmapped or moving obstacles. The proposed planning system is integrated into a complete autonomous quadrotor with global and local perception and localization sub-modules. Our system is validated by performing aggressive flights in challenging indoor/outdoor environments. We release all components in our quadrotor system as open-source ros-packages.

• The paper introduces a system that converts human-piloted trajectories into smooth, kinodynamically feasible paths using convex polyhedral corridors.
• It employs coordinate descent spatial-temporal optimization to enhance trajectory efficiency while reducing energy consumption and travel time.
• The system integrates real-time local re-planning with ESDF and B-spline methods, ensuring robust and adaptive flight in dynamic, obstacle-rich environments.

Overview of the Teach-Repeat-Replan System for Quadrotors

The paper "Teach-Repeat-Replan: A Complete and Robust System for Aggressive Flight in Complex Environments" presents a comprehensive system designed to enhance the operational capabilities of autonomous quadrotors, specifically targeting applications that demand aggressive and precise flight in environments with significant obstacles. Leveraging a pivotal improvement over traditional teach-and-repeat frameworks, this system introduces automated trajectory smoothing and an adaptive re-planning mechanism, providing an enhanced robustness to environmental changes and dynamic obstacles.

Methodology and System Architecture

The paper introduces a novel approach to the teach-and-repeat mechanism by focusing on converting human-piloted trajectories into topologically equivalent, smooth, and kinodynamically feasible paths. The methodology incorporates a polyhedral flight corridor construction that ensures safety and offers flexibility in path optimization. The authors employ an advanced convex cluster inflation algorithm to generate these corridors, significantly extending the available optimization space compared to conventional axis-aligned constriction methods.

Key Components:

1. Convex Polyhedron Flight Corridor: This is a critical concept that replaces traditional cube-based corridors. The method inflates free space around a teaching trajectory using convex polyhedral clusters to capture maximal free space, thus enhancing trajectory flexibility.
2. Coordinate Descent Spatial-Temporal Optimization: The optimization framework decouples the problem into spatial and temporal components. It iteratively optimizes the trajectory to minimize energy while respecting the physical dynamics of the quadrotor and leveraging the flexibility afforded by the convex corridors.
3. Local Re-planning for Dynamic Environments: Addressing the need for resilience in changing settings, the system integrates real-time perception updates and re-planning. Utilizing local mapping via incremental Euclidean Signed Distance Field (ESDF), the system performs obstacle-aware local trajectory modifications using a B-spline optimization method.

Numerical Results and Performance

The paper substantiates its claims with extensive simulations and real-world tests, demonstrating significant improvements over prior approaches in various indoor and outdoor environments. Numerical results indicate reduced trajectory lengths, faster completion times, and minimized energy usage compared to baseline methods, emphasizing the enhanced efficiency and robustness of the proposed system.

Notable Findings:

• Trajectory Efficiency: The new approach shows superior efficiency in trajectory pathfinding and execution compared to gradient-based and waypoint-based methods, largely attributable to the increased optimization space from the polyhedral corridor and the coordinate descent optimization.
• Robustness Under Adverse Conditions: The inclusion of a sliding-window re-planning mechanism ensures safety and adaptability during flight, allowing the quadrotor to navigate effectively amid unpredictable environmental shifts and moving obstacles.

Practical Implications and Future Directions

This research positions the proposed teach-repeat-replan system as a valuable asset for operations in cluttered and dynamically changing environments, offering a template for future developments in autonomous aerial navigation. The open-source release encourages further innovations and adaptations by the broader community.

Future advancements could explore more generalized applications beyond the quadrotor context, potentially refining the adaptation mechanisms for a wider array of aerial and non-aerial vehicles. Moreover, further integration with machine learning models could enhance the predictive and decision-making capabilities of the system, pushing the boundaries of autonomy in robotic systems.

In conclusion, the paper demonstrates meaningful progress in trajectory generation and execution for autonomous quadrotors, promising enhanced safety, efficacy, and usability in challenging operational scenarios.