TL;DR
This paper introduces a novel hybrid quadruped-quadrotor robot (QQR) capable of both ground crawling and aerial flight. They developed a hierarchical control system that allows seamless transitions between these modes, enabling the robot to navigate complex and dynamic environments.
What Problem Does This Paper Address?
Robots are increasingly expected to operate in diverse and challenging environments. While quadrupeds excel at traversing rough terrain and offer long endurance, they are limited in speed and ability to overcome large obstacles. Quadrotors, on the other hand, offer high mobility and speed but have limited flight times. This paper addresses the need for a single robotic platform that can combine the strengths of both, enabling it to perform complex missions requiring both ground-based endurance and aerial agility, especially in unstructured and dynamic environments where switching between locomotion modes is crucial.
Methods
The core of this work is the development of a hybrid quadruped-quadrotor robot (QQR). This robot integrates a micro, lightweight quadruped platform with a quadrotor system.
The control strategy is hierarchical, consisting of:
- Low-Level Locomotion Control:
- Terrestrial: A bio-inspired, Central Pattern Generator (CPG)-based controller is used to generate various quadruped gaits (walk, trot, translation, rotation). This controller utilizes a two-layer CPG network (thigh and shin layers) to manage the eight leg joints and adapt to different terrains.
- Aerial: A model-based aerial locomotion controller generates different quadrotor configurations (X, O, H, T) for various flight maneuvers. A key innovation here is a geometry-aware compensation controller that modifies rotor speeds to counteract thrust loss during morphing (changing aerial configurations).
- High-Level Motion Control: A finite-state machine (FSM) orchestrates the transitions between terrestrial and aerial modes. This controller uses onboard sensors (stereo camera, IMU, optical flow) to assess the environment and dynamically switch locomotion strategies. It also incorporates a reward function for optimizing terrestrial movement and uses experimental data to quantify aerial morphing compensation.
Datasets
Not discussed in the paper. The paper focuses on simulation and experimental validation of the proposed robot and control system.
Key Insights and Ideas
- Synergy of Locomotion Modes: The QQR demonstrates that combining quadrupedal crawling with quadrotor flight offers significant advantages over single-mode robots, particularly for complex missions in varied environments.
- Bio-Inspired Terrestrial Control: The use of CPGs for terrestrial locomotion provides a robust and adaptable way to generate natural-looking gaits, allowing the robot to handle different ground surfaces and conditions.
- Adaptive Aerial Morphing: The geometry-aware compensation controller is crucial for maintaining stable flight while the robot changes its aerial configuration (morphing). This allows for more agile maneuvers in confined spaces.
- Hierarchical Control for Seamless Transitions: The hierarchical control architecture effectively manages the complexity of switching between vastly different locomotion modes, enabling the robot to adapt to dynamic environmental changes.
- Environmental Adaptability: The QQR's ability to autonomously select appropriate locomotion modes (e.g., trot on smooth ground, walk on rough terrain) highlights its practical utility in real-world scenarios.
Key Results
- Terrestrial Locomotion: The CPG-based controller successfully generated various locomotion patterns (walk, trot, translation, rotation) and demonstrated adaptive behavior on different terrains (smooth vs. rough). For instance, trot was faster on smooth ground, while walk was preferred on rough terrain for better reliability.
- Aerial Morphing and Tracking: The aerial compensation controller showed improved position tracking performance during morphing maneuvers. For example, on a circular path, the error was reduced to less than 0.1m. The robot could successfully transition between X and H configurations to navigate narrow spaces.
- Terrestrial-Aerial Transition: Experimental results showed successful transitions between crawling and flying modes to overcome obstacles. The robot could crawl over a board obstacle and then fly over a box obstacle, demonstrating efficient power utilization and higher traversal rates.
- Endurance: The QQR achieved an aerial endurance of approximately 10 minutes (covering ~2000m) and a terrestrial endurance of up to 30 minutes (covering ~150m).
Limitations and Future Work
- Limitations:
- The paper acknowledges that the current system is a prototype and further optimization is needed.
- While the control scheme is effective, the paper implies that fine-tuning of parameters for optimal performance in diverse environments is an ongoing process.
- The paper mentions that the "loss rate of terrestrial locomotion" and "loss rate of aerial locomotion" are quantified, suggesting that some performance degradation occurs during transitions or specific maneuvers, though the compensation controller aims to minimize this.
- Future Work:
- The authors plan to focus on enhancing autonomous navigation for intelligent motion planning, incorporating onboard sensing and computing for more sophisticated decision-making in demanding terrains and real-world applications.
- Further research into optimizing the trade-off between flight time and crawling time for specific tasks is implied.
- The paper suggests that more terrestrial locomotion patterns can be mimicked and developed via CPG-based controllers to further bridge the gap between robotic and biological locomotion.
Summary updated