Mobile Robots

Robots move around to accomplish useful tasks, and often we want them to outperform humans in specific tasks. In our lab, we strive to achieve a new mode of robotic mobility based on innovative concepts, with the goal of enabling new potential applications for robots.

Multi-modal locomotion robots

In this research, our aim is to extend the mobility capabilities of robots beyond those currently available by creating robots that utilize at least two different means of movement. Indeed, numerous mobile robots specialize in one form of locomotion, whereas very few robots can perform complex locomotion tasks that combine multiple locomotion methods. A robot equipped with multiple locomotion mechanisms can not only traverse challenging environments where a single-modal locomotion robot can have difficulty, but also optimize its movement by switching between the possible means of moving depending on the environmental conditions. In addition, the distinct locomotion mechanisms have the potential to make a synergy when acted together, resulting in the facilitation of highly intricate locomotion. These innovative multi-modal locomotion robots have the potential to facilitate new robotic applications, thereby expanding the role of mobile robots in our society.

A representative example of such a multi-modal locomotion robot is LEONARDO, or LEO for short, which is an acronym of LEgs ONboARD drOne. This robot has two different locomotion mechanisms, that is, multi-joint legs and propeller-based thrusters, thereby achieving both terrestrial and aerial locomotion as well as the transition between walking and flying. The goal of LEO is twofold: (i) to enable robotic locomotion capabilities by leveraging its multi-modality of flying and walking, and (ii) to study the underlying robot design, dynamics, and control challenges of such a hybrid robotic platform, especially at the interface between walking, takeoff, and landing by using synchronous control of propellers and articulated joints. By leveraging the advantages of the two locomotion mechanisms, LEO specializes in walking with delicate balance, enabling it to traverse a slack rope, ride a skateboard, or resist external disturbances. We are currently investigating on how to further enhance the locomotion capability of such a hybrid robot, with the eventual goal of developing it into a robotic worker that could assist or replace human workers in dangerous environments.

Collaborative multi-robot systems

Much like humans, robots can enhance their performance when they collaborate with one another. Collaboration among robots can be advantageous in various scenarios, such as search and rescue missions, warehouse automation, and manufacturing. When robots collaborate, they can distribute tasks, share information, and complement each other's capabilities, leading to improved efficiency and performance. In our lab, we are actively researching approaches to integrate multiple robots into a system and elevate the practicality of multi-robot systems in real-world applications. 

One major research direction in our lab is to develop a system of multirotors that fly together for package delivery. The idea of delivering a package using aerial vehicles for daily purposes has been widely proposed for the past decade. However, most of the existing approaches mainly rely on the concept of using a single vehicle, which significantly limits the type, size, and weight of the package that could be delivered through the air. In this research, we aim to develop a standardized set of multirotors that can operate in teams and handle a wide range of payloads for delivery. We achieve this by introducing a varying number of vehicles depending on the need and by adopting innovative adhesive material that can be easily attached to and detached from rough or wet surfaces of the payload.

Meanwhile, collaboration between multiple robots can happen not only among robots of the same kind but also in a system of heterogeneous robots. Robots operating in different environments can complement each other and overcome the limitations of individual robots when acting as a team. One research project currently underway in our lab involves using aerial and ground robots together to guide and evacuate large crowds of people in emergency situations. In this scenario, aerial robots can observe a wide area and quickly gather information to share with ground robots, which can then interact with people and direct them to safety. To ensure effective and efficient evacuation, a reinforcement learning approach could be employed to train the robots' actions under various scenarios beforehand, enabling them to choose the best course of action in real emergency situations.

Tensegrity robots

Tensegrity structures are constructed with isolated rigid rods connected by a network of elastic cables providing tension to hold the structures. By delicately balancing the cable tension forces and rod compression forces, the structure is able to maintain its shape without collapsing. The overall shape of the structure is determined by the distribution of internal forces across its members. If the internal forces can be manipulated using additional actuators, the structure's shape can be changed in a controlled manner, resulting in a tensegrity robot with shape-shifting ability. This robot is capable of movement by leveraging its shape deformation effectively.

Previously, we studied two locomotion methods for spherical tensegrity robots: rolling and hopping. As a naturally stable structure, the robot rolls by deforming its outer shape in a way that moves its center of gravity projection outside the supporting polygon on the ground, resulting in punctuated rolling. Due to the close coupling of internal forces across the robot's structural elements, a small force change in any tensile member results in structural-level deformation, making it challenging to predict how the robot's shape will change. This is known as the forward problem of a tensegrity structure, which we address by building upon the dynamic relaxation technique commonly used in tensile structure analysis. The inverse problem involves finding the optimal shape and associated control inputs for rolling, which we solve using a multi-generation Monte Carlo approach.

Hopping, on the other hand, is a unique mode of locomotion applicable to tensegrity robots. Spherical tensegrity robots typically have an empty space at their center where a compressed-air thruster may be located to enable hopping. This is possible because tensegrity robots are light enough to hop against gravity and can also withstand landing impact shock thanks to their structural-level compliance. A preliminary study conducted with NASA researchers showed that a spherical tensegrity robot could be deployed for long-range exploration of the Moon by utilizing hopping and rolling locomotion methods.

Building upon our previous research, our current focus is on tensegrity robots with more general shapes and advanced locomotion capabilities.